Method and device for accurate positioning of objects

Abstract

A object’s position is determined using a cloud of timestamped points, e.g. obtained via LIDAR. A set of points representing the object is obtained (S310) among the point cloud, and a subset of points

Description

FIELD OF THE INVENTION The present invention generally relates to localization and mapping systems. More specifically, the present invention relates to a method and device for accurate positioning of objects using a cloud of points generated by a LIDAR. BACKGROUND OF THE INVENTION LIDAR sensors, which stands for Light Detection and Ranging, possess the capability to generate what is commonly referred to as a “point cloud”, by capturing and processing an extensive collection of individual data points in a three-dimensional space. These data points are precise measurements that represent the distances from the LIDAR to objects within the sensor’s range. When these data points are combined, they create a comprehensive representation of the environment surrounding the LIDAR. For instance, these data points enable to clearly visualize the shapes and edges of the various objects present in the surrounding environment. The remarkable capabilities of LIDARs to efficiently generate a detailed point cloud have generated major interest in a wide variety of industries ranging from autonomous vehicles to environmental monitoring. From an operational standpoint, a LIDAR is equipped with a laser used to target objects or surfaces within the sensor’s range using laser pulses, before measuring the time for the reflected laser pulses to return to the sensor. Also, a LIDAR can function in a fixed direction or undertake scanning across multiple directions. In this latter operating mode, the direction targeted by the laser changes according to a scan pattern specifying how the LIDAR’s field of view, i.e., direction targeted by the laser, is systematically covered. Scan patterns can vary and may include horizontal and / or vertical sweeps, multi-beam arrangements, or combination thereof. When a LIDAR is equipped with a single laser, which is typically the case, and is used to generate a point cloud of a surrounding environment, the points of the generated point cloud are acquired at different time instances, also referred to as “timestamps”. This is because a significant time is required to change the direction targeted by the laser equipping the LIDAR and / or to separate two consecutive laser pulses emitted in the same direction. In various fields where a detailed representation of the surrounding environment is required, the use of LIDARs has become essential, especially when safety and high precision are imperative. In the field of autonomous vehicles, for example, the use of LIDARs in conjunction with multiple other types of sensors, including cameras and radars, has become indispensable to clearly perceive the surrounding environment, even in difficult perception conditions. The main objectives of the deployment of LIDARs are the detection of presence of other users and objects, e.g., pedestrians and other vehicles, and accurately localizing their positions. In an intelligent transport system (ITS), fixed and mobile ITS stations equipped with LIDARs are used to perceive the surrounding environment and to share, using dedicated ITS messages, information about the perceived objects with other ITS users. These dedicated ITS messages, such as CPMs, are generally generated and transmitted periodically. Since the time instance at which an object is perceived may not coincide with the time instance at which the reporting ITS message is generated and transmitted, it is common to report a perceived object by providing a position and an associated time instance indicating when the provided position is measured. When both the LIDAR and a perceived object are fixed, the LIDAR generates a point cloud precisely representing the perceived object, although the data points of the point cloud are measured at different timestamps. This precise representation may be used to determine a position and an associated time instance for the perceived object by selecting a data point among those representing the perceived object. In this case, the associated time instance may merely be the timestamp associated with the selected data point. However, this selected data point may belong to any part of the object, making this approach incompatible with high-precision applications, such as ITS. However, if the perceived object, the LIDAR or both move while the LIDAR is scanning its surrounding environment, the data points of the point cloud that represent the perceived object are measured at different timestamps and for different positions (with respect to the LIDAR) of the perceived object. Also, a deformation of the perceived object as represented in the point cloud, referring to the change in the shape, size or structure of the perceived object, is likely to occur. For instance, if the LIDAR scan pattern and the perceived object are moving in a same direction or in opposite directions, an increase or a decrease, respectively, in the size of the perceived object as represented in the point cloud can be observed. In this case where the LIDAR and / or the perceived object are moving, merely selecting a data point among those representing the perceived object is not an effective solution, since these data points do not represent precisely the perceived object. In order to deal with the deformation of a perceived object in a point cloud and be able to provide a precise position of a perceived object, it has been reported, in many state-of-the-art documents, to proceed by reconstructing the perceived object in the point cloud. To do so, processing on all the data points of the point cloud that represent the perceived object is to be performed. Also, additional information about the perceived object, such is its speed, may be mandatory to successfully reconstruct the perceived object. Such additional information may be retrieved using multiple point clouds recorded by the LIDAR and representing all the perceived object. However, entirely reconstructing the perceived object is a processing and time-consuming task requiring advanced algorithms and significant computational resources, since the number of data points to be processed is huge and may exceed tens of thousands. Also, requiring multiple point clouds to conduct the reconstruction task makes these state-of-the-art solutions requiring a significant processing time, making them incompatible with many real-time applications, such as ITS. Thus, there is a need for an improved method for determining a precise position and an accurate associated time instance fora fixed or mobile object appearing in a cloud of data points. SUMMARY In accordance with a first aspect of the invention, there is provided a method for determining a position for an object using a cloud of timestamped points, the method comprising: o obtaining a set of points, among the cloud of points, representing, at least partially, the object; o selecting, from the set of points, at least one subset of points; o determining a spatial position and an associated time instance for the object, the spatial position and the associated time instance are determined based on the at least one subset of points. In some embodiments, selecting, from the set of points, at least one subset of points may comprise: selecting, from the set of points, at least one subset of points, each of the at least one subset of points being associated with a spatio-temporal selection extent less than a spatio-temporal extent of the obtained set of points. In some embodiments, selecting, from the set of points, at least one subset of points may comprise: selecting, from the set of points, at least one subset of points, each of the at least one subset of points being associated with a spatial selection extent less than a spatial extent of the obtained set of points. In some embodiments, selecting, from the set of points, at least one subset of points may comprise: selecting, from the set of points, at least one subset of points, each of the at least one subset of points being associated with a temporal selection extent less than a temporal extent of the obtained set of points. Accordingly, instead of handling an entire cloud of points whose size, in terms of number of points, may exceed tens of thousands, requiring significant memory and processing resources, and where points representing an object may be close or coincide, whereas they are distant in time and / or space, the various aspects and embodiments of the invention overcome these limitations by handling a few size-limited subsets of points selected, from the cloud of points, based on spatial and / or temporal bases. The points of each subset exhibit an enhanced spatial and / or temporal coherence, with respect to the cloud of points. In this way, a comprehensive overview, in both space and time, of an object, appearing in the cloud of points, is obtained, making it possible to determine a precise spatial position and an accurate associated time instance for the object. In some embodiments, the selection extent may be less than the ratio of the corresponding extent of the obtained set of points to the number of the selected subsets of points. In some embodiments, selecting, from the set of points, at least one subset of points may further comprise: o determining a bounding-box enclosing the obtained set of points; o identifying an orientation for the object; o selecting, as a first subset of points, a first group of points that are closer to a first face of the bounding-box substantially perpendicular to the identified orientation than to a second face of the bounding-box that is opposite to the first face, and selecting, as a second subset of points, a second group of points that are closer to the second face of the bounding-box than to the first face. In some embodiments, selecting, from the set of points, at least one subset of points may further comprise: selecting, as a single subset of points, a group of points that are representative of the whole object. In some embodiments, selecting, from the set of points, at least one subset of points may further comprise: selecting, as a single subset of points, a group of points that are representative of a specific part of the object. In some embodiments, the cloud of timestamped points may be obtained using a LIDAR executing, at least once, a LIDAR scan pattern. In some embodiments, selecting, from the set of points, at least one subset of points may further comprise: o determining an approximate position for the object; o identifying a plurality of scan configurations enabling the LIDAR to scan in the direction corresponding to a region of interest centered at the approximate position of the object; o selecting, as a single subset of points, a group of points that are obtainable using identified scan configurations. In some embodiments, determining a spatial position and an associated time instance for the object may comprise, for each of the at least one subset of points: o determining an approximate spatial position for the object, based on positions associated with the points of the subset; o determining an approximate time instance for the object, based on timestamps associated with the points of the subset. In some embodiments, the spatial position of the object may be determined as a barycenter of the determined at least one approximate spatial position, and the associated time instance may be determined by averaging the determined at least one approximate time instance. In accordance with a second aspect of the invention, there is provided a method of communication in an Intelligent Transport System (ITS), comprising: transmitting an ITS message comprising items of information describing an object of interest, the items of information include a spatial position and an associated time instance, the spatial position and the associated time instance are determined according to any aspect or embodiment described above. In accordance with a third aspect of the invention, there is provided a processing device configured to perform the method according to any aspect or embodiment described above. Any feature in one aspect of the invention may be applied to other aspects of the invention, in any appropriate combination. In particular, method aspects may be applied to apparatus / device / unit aspects, and vice versa. Furthermore, features implemented in hardware may be implemented in software, and vice versa. Any reference to software and hardware features herein should be construed accordingly. For example, in accordance with other aspects of the invention, there are provided a computer program comprising instructions which, when the program is executed by a processing unit, cause the processing unit to carry out the method of any aspect or example described above and a computer readable storage medium carrying the computer program. BRIEF DESCRIPTION OF THE DRAWINGS Embodiments of the invention will now be described, by way of example only, and with reference to the following drawings in which: Figure 1 is a schematic diagram of an Intelligent Transport System (ITS) in which the present invention may be implemented, according to one or more embodiments; Figures 2a, 2b and 2c show examples of a point cloud representing an ITS environment including a vehicle, the vehicle being stationary in Figure 2a and mobile in Figures 2b and 2c; Figure 3 is a diagram illustrating a flowchart of a method for determining a spatial position and an associated time instance for an object of interest, according to embodiments of the invention; Figure 4a schematically illustrates a two-dimensional point cloud representing a moving object of interest having three constituent parts; Figure 4b schematically illustrates a subset of data points, extracted from the point cloud of Figure 4a, having a spatial extent corresponding to a specific constituent part of the object of interest; Figure 4c schematically illustrates a subset of data points, extracted from the point cloud of Figure 4a, having a temporal extent corresponding to a time period less than that required by the LIDAR to execute, once, a LIDAR scan pattern; Figure 4d schematically illustrates three subsets of data points, extracted from the point cloud of Figure 4a, having different spatio-temporal extents; Figures 5a and 5b show subsets of data points having different spatio-temporal extents and representing two different constituent parts of a moving object of interest; Figures 6 is a diagram illustrating a flowchart of a method for selecting multiple subsets of data points from a point cloud, according to some embodiments of the invention; Figures 7 is a diagram illustrating a flowchart of a method for selecting a single subset of data points from a point cloud, according to other embodiments of the invention; Figures 8 is a diagram illustrating a flowchart of a method for processing a point cloud representing, at least partially, an object of interest, according to embodiments of the invention; Figure 9 is a schematic diagram of a processing device in which methods according to embodiments of the present invention may be implemented. DETAILED DESCRIPTION OF EMBODIMENTS Embodiments of the present invention provide methods, devices and computer program products for determining a spatial position and an associated time instance for a fixed or mobile object appearing, at least partially, in a cloud of timestamped data points generated by a LIDAR, for example. Embodiments of the present invention provide a precise spatial position and an accurate associated time instance by processing one or more subsets of data points representative of the object, the data points of each of the one or more subsets being selected from the cloud of data points in a way to satisfy certain spatial, temporal or spatio-temporal selection conditions. By processing multiple subsets of data points, instead of processing the entire cloud of data points, a spatial position and an associated time instance for the object are determined with a reduced computational complexity. Methods, devices and computer program products according to the various embodiments of the present invention may be implemented in surveillance systems comprising at least one LIDAR configured to acquire a cloud of timestamped data points of an area of interest, the LIDAR being calibrated relative to a reference frame and being configured to operate according to a LIDAR scan pattern. The surveillance system may be part of an intelligent transport system (ITS) where multiple LIDARs are deployed along roads and at intersections in order to detect congestion and notice accidents, for example. The surveillance system may be part of an autonomous vehicle where multiple LIDARs are used to perceive the surrounding environment, especially to detect obstacles and to identify lanes and traffic signals. The surveillance system may be part of a public surveillance system used in areas requiring additional security or ongoing monitoring in order to prevent crimes and monitor the flow of crowds, for example. For illustration purposes only, the following description will be made with reference to an ITS comprising a fixed or mobile LIDAR calibrated relative to a reference frame and configured to acquire a cloud of timestamped points of an area of interest comprising an object, such as a vehicle or a pedestrian, according to a LIDAR scan pattern. However, the skilled person will readily understand that the various embodiments of the invention apply in other types of surveillance systems. Generally, the invention may be integrated in any surveillance system deploying LIDARs. In the following description, the expressions “cloud of timestamped data points”, “cloud of data points” and “point cloud” are used interchangeably to designate a three-dimensional representation of an environment or a specific region generated using a LIDAR and composed of an extensive collection of individual data points. Each of these data points represents, i.e., comprises, a precise measurement of the distance between the LIDAR and the target object or surface being detected. In addition to the distance, a data point may comprise supplementary details such as attenuation, frequency shift and phase shift. These characteristics reflect the alterations experienced by the emitted optical signal upon its return to the LIDAR and may provide additional information about the target object or surface being detected. In addition, for each point of the point cloud, a timestamp is associated with it. This timestamp provides information about the exact time when the LIDAR measurement associated with the data point was taken. Without limitation, a timestamp may represent the time instance at which the LIDAR measurement associated with the data point is triggered. In the following description, the expressions “LIDAR scan pattern”, “scan pattern” and “LIDAR pattern” are used interchangeably to designate how the LIDAR’s field of view, i.e., direction targeted by the laser equipping the LIDAR, evolves over time and in space. In other words, the LIDAR scan pattern specifies the time instance and the laser-targeted direction according to which each LIDAR measurement is conducted. Scan patterns can vary and may include horizontal and vertical sweeps, rotations, multi-beam arrangements, or combination thereof. An example of a LIDAR scan pattern is 360-degree scanning where the LIDAR is designed to rotate continuously, creating a full 360-degree scan of its surrounding environment. More generally, the LIDAR scan pattern may be a custom scan pattern where the LIDAR is programmed to follow a specific scan pattern based on specific application requirements. For example, the LIDAR may focus more on scanning roads and intersections in an ITS application. When the LIDAR completes a scan according to a LIDAR scan pattern, the obtained data points represent what is commonly referred to as “a frame of data points”. Of course, a cloud of data points may comprise multiple frames of data points. In the following description, the expressions “spatial extent”, “temporal extent” and “spatio-temporal extent” are used to describe sets, subsets and groups of data points belonging to a cloud of timestamped data points. The spatial extent of a set of data points refers to the overall geographic coverage or area covered by these data points. It describes the range or geographical span within which these points are distributed or located. The spatial extent of a set of data points may be determined based on the distance measurements associated with the data points and, optionally, on the supplementary details included in these data points. The temporal extent of a set of data points, on the other hand, represents the overall time span during which these data points exist or are relevant. It describes the duration or time period over which these points are observed, measured, or recorded. The temporal extent of a set of data points may be determined based on the timestamps associated with the data points. Furthermore, the spatiotemporal extent of a set of data point refers to both the geographic coverage or area and the time period during which these points exist or are relevant. It describes the combined spatial and temporal range within which these points are observed or located. Of course, the spatio-temporal extent of a set of data points may be determined based on both the distance measurements and the timestamps of the data points. By denoting by the real-valued numbers S and T the spatial extent and the temporal extent of a set of data points, respectively, the spatio-temporal extent of the set of data points may be represented by a complex-valued number whose real and imaginary parts are S and T, respectively. In the following description, a cloud of timestamped data points generated using a LIDAR according to a LIDAR scan pattern, and representing an environment comprising one or more objects is considered, one of the one or more objects being an object of interest to be localized by determining a spatial position and an associated time instance. The LIDAR may either be fixed or mobile. The object of interest, appearing at least partially in the point cloud, may also be either fixed or mobile with respect to the LIDAR. Referring to Figure 1, there is shown an example of an ITS 100 in which embodiments of the present invention may be implemented. The ITS 100 comprises a LIDAR 110 configured to measure and record a cloud of timestamped data points of a surrounding environment that comprises multiple objects, at least one of the objects being an object of interest 120. The LIDAR 110 may be configured to measure the point cloud according to a LIDAR scan pattern that may be, without limitation, a 360-degree scan pattern or a custom scan pattern. The use of a 360-degree scan pattern is particularly beneficial when the object of interest 120 is likely to be anywhere in the ITS 100, such as a pedestrian who can be on the road or on the sidewalk. The use of a custom scan pattern, on the other hand, is more compatible with scenarios where the object of interest 120 is compelled to move through specific areas, such as a vehicle that can only move on the road. The LIDAR scan pattern also specifies the way the LIDAR’s field of view evolves over time. For example, a 360-degree scan pattern may be executed by utilizing two nested loops corresponding, respectively, to horizontal and vertical sweeps. More generally, a LIDAR scan pattern may be executed gradually, i.e., step-by-step progression, or at random. Furthermore, the LIDAR 110 may be fixed or mobile. In the latter case, the LIDAR 110 may be mounted on a vehicle, for example. In the case of a fixed LIDAR, the frames of data points, obtained when the LIDAR executes, multiple times, the LIDAR scan pattern, may all be representative of a same environment, in particular multiple data points, each belonging to a respective frame of data points and being associated with a different timestamp, are associated with a same LIDAR’s field of view. In the case of a mobile LIDAR, each frame of data points may be representative of a different environment. The ITS comprises multiple objects, including an object of interest 120, that can be classified depending on various factors or criteria, including but not limited to their size, shape, or mobility status. For example, an object may be classified as either fixed or mobile. The embodiments of the present invention are compatible with the following scenarios: both the LIDAR 110 and the object of interest 120 are either fixed or mobile, the LIDAR 110 is fixed and the object of interest 120 is mobile, and the LIDAR 110 is mobile and the object of interest 120 is fixed. The embodiments of the present invention determine a spatial position and an associated time instance for the object of interest 120, if the object of interest 120 appears, at least partially, in a cloud of timestamped data points generated by the LIDAR 110. The object of interest 120 may appear in one or more frames of data points composing the point cloud. The ITS further comprises a processing unit 130 that may be incorporated in an ITS station (not represented in Figure 1) capable of transmitting and receiving ITS messages within the ITS 100. The processing unit 130 is configured to receive, from the LIDAR 110, a cloud of timestamped data points representing an environment comprising an object of interest 120. The data points may be received each time a data point is recorded, each time a frame of data points is recorded, or each time the entire cloud of data points is recorded. The processing unit 130 is further configured to process the point cloud in order to determine a spatial position and an associated time instance for the object of interest 120. To do so, the processing unit 130 implements methods according to various embodiments of the present invention. The ITS also comprises a transmission unit 140 that is incorporated in the ITS station. The ITS station may be fixed by equipping a Road Side Unit (RSU), or mobile by equipping a vehicle ora Vulnerable Road User, VRU. The transmission unit 140 is configured to generate and transmit ITS messages comprising perception information about the object of interest 120. The perception information includes, but not limited to, an assigned identifier, a mobility status, a spatial position and an associated time instance. The ITS messages may further comprise sensor information about the deployed LIDAR that is used to determine the spatial position and the associated time instance. Examples of ITS messages that may be generated and transmitted by the ITS station includes Collective Perception Message (CPM), Cooperative Awareness Messages (CAMs), Vulnerable Road Users Awareness Messages (VAMs), and Decentralized Environmental Notification Messages (DENMs). Referring to Figures 2a, 2b and 2c, there is shown, in each of these Figures, an example of a cloud of timestamped data points corresponding to a single frame of data points and representing an ITS environment including a vehicle as an object of interest, the point cloud being acquired using a fixed LIDAR. The LIDAR executes a 360-degree scan pattern comprising horizontal and vertical sweeps, a vertical sweep being executed at each horizontal scan angle. While the object of interest shown in Figure 2a is fixed, the objects of interest of Figures 2b and 2c are mobile and move, with a different speed, in a same direction and in an opposite direction, respectively, when compared to the horizontal scan direction of the LIDAR. Due to the fact that the object of interest is moving and that the data points are measured at different timestamps, a deformation (i.e., a distortion) in the shape of the object of interest resulting in an increase or in a decrease of its size is clearly visible in Figures 2b and 2c. The deformation in the shape of the object of interest may be even more pronounced with other types of LIDAR scan patterns, or when the point cloud is composed of multiple frames of data points. The deformation in the shape of the object of interest observed in Figures 2b and 2c also appears in the scenario when both the LIDAR and the object of interest are moving, or when the LIDAR is moving and the object of interest is fixed. With advances in technology, detection algorithms based on artificial intelligence and machine learning have become increasingly capable of autonomously performing tasks such as detecting, identifying and approximately positioning objects appearing in a cloud of timestamped data points. A large majority of these detection algorithms provide, as output for each detected object, a three-dimensional bounding-box delimitating the data points representing the detected object. The provided bounding-box can then be used to infer a spatial position and an associated time instance for the delimitated detected object. For instance, the inferred spatial position may correspond to the center of the bounding-box or may be the projection onto the ground of the center of the bounding-box. The inferred associated time instance, referring to the time instance when the detected object was precisely at the inferred spatial position, corresponds generally, according to state-of-the-art detection algorithms, to a timestamp of a data point determined as representing, i.e., belonging to, the detected object, or corresponds to the beginning, the middle or the end of a frame of data points, composing the point cloud, where the detected object appears for the first time. However, as illustrated in Figures 2b and 2c, a deformation in the representation of an object of interest within the point cloud is likely to occur, leading, for instance, to a bounding-box with incorrect dimensions. This results in the inferred spatial position of the object of interest being only approximate. Also, the spatial position and the associated time instance are generally determined, according to state-of-the-art detection algorithms, separately and using different approaches and techniques, which makes the inferred spatial position barely correspond to the inferred associated time instance. Figure 3 is a flowchart of a method 300 for determining a spatial position and an associated time instance for an object of interest appearing in a cloud of timestamped data points. The point cloud may be composed of any number of frames of data points. The point cloud may be acquired using a LIDAR executing any type of LIDAR scan pattern. The LIDAR may be fixed or mobile, and the object of interest, whose motion is assumed to be independent of that of the LIDAR, may also be fixed or mobile. The method 300 may be implemented at any processing device, such as the processing unit 130 described with reference to Figure 1. At step S310, a set of data points representing, at least partially, the object of interest is obtained from the cloud of timestamped data points. This may comprise applying an object detection process to identify some or all of the data points within the point cloud that represent the object of interest. Generally, an object detection process may be implemented differently depending on the type of the object of interest (e.g., vehicles, trees, buildings) and the characteristics of the point cloud (e.g., resolution, point density). An object detection process may comprise a step of segmentation to segment the point cloud into meaningful clusters or segments, a step of feature extraction to extract relevant features from the segmented clusters, and a step of applying an object detection algorithm to the segmented clusters and extracted features. The object detection algorithm may be machine learning-based where a machine learning model (e.g., deep neural networks like CNNs or PointNet) is trained to identify the object of interest within the clusters and / or segments. The object detection algorithm may also be geometry-based where geometric properties such as size, shape and orientation are used to recognize the object of interest. In this way, the objection detection process determines whether a data point of the point cloud represents, i.e., belongs to, the object of interest, according to a relevance score. For example, the obtained set of data points may include all the data points identified as representing the object of interest. At step S320, one or more subsets of data points are selected from the obtained set of data points. When multiple subsets are selected in step S320, the size of each of the subsets, in terms of number of composing data points, may preferably be less than the ratio of the size of the obtained set of data points to the number of the selected subsets. Also, the multiple subsets may preferably be selected in such a way that there is no intersection between them. This means that a data point of the point cloud belongs to at most one subset. When only one subset is selected in step S320, the size of the subset may preferably be several orders of magnitude less than that of the obtained set of data points, and the subset may be representative of the whole object of interest or of a specific part of the object of interest. In this latter case of a single subset being selected, a downsampling may be applied to the obtained set of data points in order to select the subset. Generally, the number of the selected subsets may be chosen depending on features of the object of interest, including but not limited to the size, the shape and the mobility status. For example, a single subset may be selected when the object of interest is fixed, two subsets may be selected when the object of interest moves at a constant speed, and three or more subsets may be selected when the object of interest accelerates. Also, some characteristics of the LIDAR, such as its mobility status and its LIDAR scan pattern, may be taken into consideration when determining the number of the selected subsets. In a first variant of step S320, each subset is selected in such a way that its spatial extent, i.e., the spatial extent of its data points, denoted s / , is less than the spatial extent of the obtained set of data points, denoted S. Each subset may extend over the entire temporal extent of the obtained set of data points. In other words, considering a moving object of interest having several constituent parts and appearing in several consecutive frames of data points composing the point cloud, a subset according to this first variant of step S320 may include all the data points that belong to only one of the constituent parts within all the frames of data points. Preferably, the spatial extent of each subset may be less than the ratio of the spatial extent of the obtained set of data points to the number of the selected subsets. More generally, the spatial extent of each subset may be less than a predetermined spatial extent threshold. When multiple subsets are selected in step S320, the subsets may be selected such that there is no intersection, in space, between them and they are distributed, uniformly or non-uniformly, across the entire spatial extent of the obtained set of data points. In this way, a reliable spatial representation of the object of interest is obtained using multiple subsets, each of the multiple subsets consisting of a limited number of data points in comparison to the size the obtained set of data points. This makes it possible to effectively deal with any deformation of the object of interest, as represented in the point cloud, by filtering out, i.e., excluding, from a subset corresponding to a part of the object of interest one or more intrusive data points representing the object of interest but corresponding to another part of the object of interest. In a second variant of step S320, each subset is selected in such a way that its temporal extent, i.e., the temporal extent of its data points, denoted ti, is less than the temporal extent of the obtained set of data points, denoted T. Each subset may extend over the entire spatial extent of the obtained set of data points. In other words, considering a moving object of interest having several constituent parts and appearing in several consecutive frames of data points composing the point cloud, a subset according to this second variant of step S320 may include all the data points that belong to all the constituent parts within only one of the frames of data points. Preferably, the temporal extent of each subset may be less than the ratio of the temporal extent of the obtained set of data points to the number of the selected subsets. More generally, the temporal extent of each subset may be less than a predetermined temporal extent threshold. When multiple subsets are selected in step S320, the subsets may be selected such that there is no intersection, in time, between them and they are distributed, uniformly or non-uniformly, across the entire temporal extent of the obtained set of data points. In this way, a reliable temporal representation of the object of interest is obtained using multiple subsets, each of the multiple subsets consisting of a limited number of data points in comparison to the size the obtained set of data points. This makes it possible to effectively deal with any deformation of the object of interest, as represented in the point cloud, by filtering out, i.e., excluding, from a subset corresponding to a temporal extent one or more intrusive data points representing the object of interest but corresponding to another temporal extent. In a third variant of step S320, each subset is selected in such a way that its spatiotemporal extent, i.e., the spatio-temporal extent of its data points, denoted hi, describing the combined spatial and temporal extent, given respectively by s / and ti, within which these data points are observed or located, is less than the spatio-temporal extent of the obtained set of data points, denoted H. The spatio-temporal extent of a subset may be considered to be less than that of the obtained set of data points if, and only if, at least one of its spatial and temporal extents is less than that of the obtained set of data points. Each subset, whose spatio-temporal extent hi= si+i*ti, may be determined as an intersection of a first subset whose spatial extent is s / and a second subset whose temporal extent is ti. In other words, considering a moving object of interest having several constituent parts and appearing in several consecutive frames of data points composing the point cloud, a subset according to this third variant of step S320 may include all the data points that belong to only one constituent part within only one of the frames of data points. Preferably, the spatio-temporal extent of each subset may be less than the ratio of the spatio-temporal extent of the obtained set of data points to the number of the selected subsets. More generally, the spatio-temporal extent of each subset may be less than a predetermined spatio-temporal extent threshold. When multiple subsets are selected in step S320, the subsets may be selected such that there is no intersection, neither in space nor in time, between them and they are distributed, uniformly or non-uniformly, across the entire spatio-temporal extent of the obtained set of data points. In this way, a reliable spatio-temporal representation of the object of interest is obtained using multiple subsets, each of the multiple subsets consisting of a limited number of data points in comparison to the size the obtained set of data points. This makes it possible to effectively deal with any deformation of the object of interest, as represented in the point cloud, by filtering out, i.e., excluding, from a subset corresponding to a spatio-temporal extent one or more intrusive data points representing the object of interest but corresponding to another spatio-temporal extent. Figure 4a schematically illustrates a two-dimensional point cloud 111 representing a moving object of interest (only the data points that represent the object of interest are shown), the object of interest having three parts indicated in the point cloud using vertical lines, dots and horizontal lines, respectively, for illustration purposes only. Of course, in real-world scenarios, the data points of the point cloud may often exhibit a same shape, making it difficult to distinguish which part of the object of interest a data point belongs to. The point cloud 111 consists of three frames of data points acquired while the object of interest is moving, as illustrated in the representation 401 of the object of interest. By focusing on the central region of the point cloud 111, it is visible that data points belonging to different parts of the object of interest overlap, rendering their processing memory- and time-consuming. In particular, the point cloud 111 includes adjacent data points (that may even coincide) that are respectively representative of different parts of the object of interest. In Figure 4b, there is shown a subset of data points 112b, extracted from the point cloud 111 of Figure 4a according to the first variant of step S320, having a spatial extent corresponding to a specific part (the front part) of the object of interest, as illustrated in the representation 402 of the object of interest. The shown subset 112b consists of three separate groups of data points, each corresponding to a different position of the specific part of the object of interest. The use of a subset of data points corresponding to a specific part of an object of interest is beneficial when the object of interest is moving and its spatial position to be determined is that of this specific part. However, this approach has limitations when groups of data points within a subset overlap. In Figure 4c, there is shown a subset of data points 112c, extracted from the point cloud 111 of Figure 4a according to the second variant of step S320, having a temporal extent corresponding to, or less than, the time period required by the LIDAR to execute, once, a LIDAR scan pattern, as illustrated in the representation 403 of the object of interest. The shown subset 112c consists of three separate groups of data points, each corresponding to a different part of the object of interest. The use of a subset of data points corresponding to a same time period is beneficial when the object of interest is fixed and its spatial position to be determined is function of its different parts. Also, this approach has limitations, especially when groups of data points within a subset overlap. In Figure 4d, there are shown three subsets of data points 112d[1 to 3], extracted from the point cloud 111 of Figure 4a according to the third variant of step S320, each of the shown subsets 112d[1 to 3] having a spatio-temporal extent corresponding to a spatial extent of a part of the object of interest and to a temporal extent equal to, or less than, a time period required by the LIDAR to execute, once, a LIDAR scan pattern. The spatial extent, the temporal extent, and therefore the spatio-temporal extent, of each of the subsets 112d[1 to 3] are different from those of the other subsets, as illustrated in the representation 404 of the object of interest. The first shown subset 112d1 may be seen as an intersection between the subsets 112b and 112c depicted in Figure 4b and 4c, respectively. Similarly, each of the other shown subset 112d[2 to 3] may be seen as an intersection between two subsets defined in a manner similar to those of Figure 4b and 4c. The use of multiple subsets of data points having different spatio-temporal extents allows to determine positions for different parts of the object of interest during different time periods. In this way, a comprehensive spatial and temporal overview of the object of interest is obtained. This is particularly advantageous when the object of interest and / or the LIDAR is mobile. Referring back to Figure 3, at step S330, a spatial position and an associated time instance are determined for the object of interest, based on the selected subsets of data points. In some embodiments associated with a single subset being selected in step S320, such as one of the subsets shown in Figures 4b, 4c and 4d, a data point of the selected subset may be picked and used to determine a spatial position and an associated time instance for the object of interest. In this case, the spatial position is determined to be the spatial position represented by the picked data point, and the associated time instance is determined to be the timestamp associated with the picked data point. More generally, the spatial position and the associated time instance may be determined based on all or a portion of the data points of the selected subset through applying one or more processing operations, such as filtering and averaging. In the case of a single subset being selected as described in the first variant of step S320 and consisting of multiple groups of data points determined in a way to have a same spatial extent, s / , each group of data points may represent a different position of the object of interest (as shown in Figure 4b), the spatial position of the object of interest may be determined as a spatial position represented by one of the groups of data points. This may be achieved by averaging the spatial positions represented by the data points of the chosen group to determine the spatial position of the object of interest. The associated time instance, on the other hand, may be determined as the timestamp of the data point that is nearest to the determined spatial position within the chosen group of data points. In the case of a single subset being selected as described in the second variant of step S320 and consisting of multiple groups of data points determined within a same temporal extent, ti, each group of data points may represent a different constituent part of the object of interest (as shown in Figure 4c), the spatial position of the object of interest may be determined as a spatial position represented by one of the groups of data points, i.e., a spatial position represented by of one of its constituent parts. This may be achieved by averaging the spatial positions represented by the data points of the chosen group to determine the spatial position of the object of interest. The associated time instance, on the other hand, may be determined as the timestamp of the data point that is nearest to the determined spatial position within the chosen group of data points. In other embodiments associated with multiple subsets being selected in step S320, such as the three subsets shown in Figure 4d, the spatial position and the associated time instance may be determined based on data points extracted from at least two separate selected subsets. Preferably, when the subsets are selected based on spatial and / or temporal bases, such as subsets selected according to the first, second or third variant of step S320, data points extracted from distant subsets, in space and / or in time, may be used to determine a spatial position and an associated time instance for the object of interest. In the case of multiple subsets being selected according to the third variant of step S320, all or some of the selected subsets may be chosen to determine a spatial position and an associated time instance for the object of interest. Generally, the selected subsets according to the third variant of step S320 have limited spatial and temporal extents, in comparison with those of the object of interest as represented in the point cloud, while reliably representing the object of interest in both space and time. In particular, the two most distant subsets, in space and in time, represents the spatial and temporal boundaries of the motion of the object of interest. For example, if the object of interest has several constituent adjacent parts and appears in several consecutive frames of data points composing the point cloud, the two most distant subsets may correspond, respectively, to a first subset, extracted from the earliest frame of data points, representing the first constituent part, and a second subset, extracted from the latest frame of data points, representing the last constituent part. This is schematically illustrated in Figures 5a and 5b where a vehicle 120, as an object of interest, is shown, along with its motion direction 500. The vehicle consists of three primary parts, corresponding to its front, middle and rear sections. While a first subset corresponding to the rear part of the vehicle and acquired within the earliest frame of data points is illustrated in Figure 5a, a second subset corresponding to the front part of the vehicle and acquired within the latest frame of data points is illustrated in Figure 5b. Relying on the two most distant selected subsets, for example, a spatial position for the object of interest may be determined as any spatial position within the spatial range delimitated by the two subsets. In this case, the spatial position of the object of interest, denoted Po, may be given by: Po=x*Pb+y*Pe, where Pb and Pe corresponding to two spatial positions respectively representative of the two subsets, and x and y are two reals. Then, the associated time instance, denoted To, may be given by: To=x*Tb+y*Te, where Tb and Te corresponding to two timestamps respectively representative of the two subsets. A spatial position representative of a subset may correspond, without limitation, to the average of the spatial positions of the data points composing the subset. Also, a timestamp representative of a subset may correspond, without limitation, to the earliest, the latest, or the average of the timestamps associated with the data points composing the subset. Alternatively, the time instance, To, may be determined first, and then the associated spatial position of the object the object of interest may be inferred based on the values, x and y, used to determine the time instance To. Of course, the number of subsets involved in the determination of the spatial position and the associated time instance, according to this approach, may be more than two subsets, especially when the LIDAR or the object of interest performs a non-uniform motion characterized by changes in speed overtime. In some embodiments where multiple subsets of data points are selected in step S320, for each of the selected subsets, an approximate spatial position for the object of interest may be determined based on spatial positions associated with the data points of the subset, and an approximate time instance for the object of interest may be determined based on timestamps associated with the data points of the subset. For example, the spatial position of the object of interest is inferred as the barycenter of the determined approximate spatial positions, and the associated time instance is inferred by averaging the determined approximate time instances. In other embodiments where a single subset of data points is selected in step S320, the subset may include the data points that contributed to detecting and / or locating the object of interest within the point cloud. These data points may be identified using artificial intelligence algorithms, such as "attention mechanism" which is commonly used to mimic how human attention works by assigning different levels of importance to different elements in an image or in a point cloud. Instead of handling an entire point cloud whose size, in terms of number of data points, may exceed tens of thousands, requiring significant memory and processing resources, and where data points representing a moving object may be close or coincide, whereas they are distant in time and / or space, the various embodiments of the invention overcome these limitations by handling a few size-limited subsets of data points selected, from the point cloud, based on spatial and / or temporal bases. The data points of each subset exhibit an enhanced spatial and / or temporal coherence, with respect to the point cloud. In this way, a comprehensive overview, in both space and time, of an object of interest, appearing in the point cloud, is obtained, making it possible to determine a precise spatial position and an accurate associated time instance for the object of interest. When multiple subsets of data points are selected according to the third variant of step S320, these subsets can be seen as samples of the combined spatial and temporal evolution of the object of interest, as appearing in the point cloud. By proceeding by interpolation, for example, it possible to determine an accurate associated time instance for each spatial position of the object of interest. Also, fora given time instance, it is possible to determine an associated spatial position of the object of interest. This flexibility in determining the spatial position and the associated time instance is allowed by the fact the same approach (subset-based approach relying on spatial and / or temporal coherence) is used to determine both of them. Figure 6 shows a diagram illustrating a sequence of substeps 600 that can be utilized to implement the step S320 of the method 300 described with reference to Figure 3, according to some embodiments of the invention. The sequence of substeps 600 is also referred to as “selection process”. This selection process 600 receives an obtained set of data points representing, at least partially, an object of interest, and provides at least two separate subsets of data points that can be utilized to determine a spatial position and an associated time instance for the object of interest. At substep S610, a three-dimensional bounding-box enclosing data points belonging to the obtained set is determined. The bounding-box may be determined in a way to encompass all the data points of the obtained set. The bounding-box may, alternatively, be determined in a way that dispersed data points are not included within it. A dispersed data point may be defined as a data point whose nearest neighbor is at a distance that exceeds a predetermined distance threshold. Preferably, the bounding-box has the geometric shape of a rectangular parallelepiped, with one face parallel to the ground on which the object of interest is either fixed or mobile. Of course, other geometric shapes for the bounding-box may be considered, such as a cylindrical bounding-box for an object of interest represented by a pedestrian. At substep S620, an orientation for the object of interest is identified. The orientation may be determined in the reference frame within which the LIDAR acquiring the point cloud is positioned. Generally, the orientation describes how the object of interest is positioned and rotated relative to the reference frame. When the object of interest is moving and the point cloud consists of multiple frames of data points, the orientation of the object of interest may be determined to points towards the data points of the last frame of data points that occupy a previously unoccupied space. When the object of interest is fixed, the orientation of the object of interest may be determined by identifying at least two constituent parts of the object of interest, relying on the obtained set of data points and / or the determined bounding-box. For example, if the object of interest is a vehicle, the front and rear constituent parts of the object of interest may align with the two smallest faces of an enclosing rectangular bounding-box. The orientation of the object of interest may also be determined using a prior knowledge of the environment where the object of interest is located. For example, the orientation of an object of interest representing a vehicle located on a one-way road may be determined to be the direction of the road. At substep S630, at least two subsets of data points are selected from the obtained set of data points, based on the determined bounding-box and the identified orientation of the object of interest. A first subset is determined in a way to include a group of data points that are closer, in terms of distance, to a first face of the bounding-box substantially perpendicular to the orientation of the object of interest than to a second face of the bounding-box that is opposite to the first face. A second subset, on the other hand, is determined in a way to include a group of data points that are closer, in terms of distance, to the second face of the bounding-box than to the first face. Preferably, the first and second subsets of data points are selected according to a same basis. For example, if the first subset is selected in a way that its spatial extent is less than a predetermined spatial extent threshold, the second subset is also selected according to the same spatial basis, i.e., to have a spatial extent less than the predetermined spatial extent threshold, the temporal extents of the two subsets may be different. In another example, if the first subset is selected in a way to have a spatio-temporal extent given by hii= sh +i*tii, sh and th being respectively the associated spatial and temporal extents, the second subset is also selected according to the same spatio-temporal basis, i.e., in a way to have a spatio-temporal extent given by hiz= si2 +i*t'i2, s / 2and femay respectively be different from sh and th, sh and s / 2 (and / or th and tiz) may be less than a predetermined spatial (and / or temporal) extent threshold. Of course, other subsets of data points within the bounding-box may also be selected, preferably according to spatial and / or temporal bases similar to those of the first and second subsets. Generally, the first and second subsets represent two distinct and extreme positions of the object of interest within the point cloud. In some embodiments compatible with a scenario of multiple subsets selected according to the selection process 600 of Figure 6, the spatial position of the object of interest may be determined as the center of the enclosing bounding-box. The associated time instance, on the other hand, is determined as the midpoint between a first timestamp representative of the first subset, such as the latest among all the timestamps of the first subset, and a second timestamp representative of the second subset, such as the earliest among all the timestamps of the second subset. This is particularly beneficial when dealing with a mobile object of interest that moves at a constant speed. In other embodiments also compatible with a scenario of multiple subsets selected according to the selection process 600 of Figure 6, the spatial position of the object of interest may be determined to be closer, in terms of distance, to one subset than to another, among the first and second subsets. In this case, the associated time instance may be determined based on a first timestamp representing the first subset and a second timestamp representing the second subset, while respecting the same proportionality as the spatial position. This is particularly beneficial when dealing with a mobile object of interest undertaking an acceleration or a deceleration. Figure 7 shows a diagram illustrating a sequence of substeps 700 that can be utilized to implement the step S320 of the method 300 described with reference to Figure 3, according to other embodiments of the invention. The selection process 700 illustrated in Figure 7 receives an obtained set of data points representing, at least partially, an object of interest, and provides a single subset of data points that can be utilized to determine a spatial position and an associated time instance for the object of interest. The selection process 700 of Figure 7 is particularly compatible with a point cloud acquired by a LIDAR executing gradually a 360-degree scan pattern, by rotating horizontally step by step in a uniform manner, and scanning vertically at each step, while adjusting the horizontal scan angle after each step. More generally, the selection process 700 of Figure 7 is compatible with any point cloud acquired by a LIDAR whose LIDAR scan pattern is precisely known beforehand. At substep S710, an approximate position for the object of interest is determined based on the obtained set of data points. This approximate position may be determined as either the center of a bounding-box enclosing data points belonging to the obtained set of data points, or as the center of any of its faces. The bounding-box may be determined as described with reference to substep S610. The approximate position may also be determined as the spatial position represented by any data point of the obtained set of data points. At substep S720, all the scan configurations of the LIDAR, defined to enable the LIDAR to scan in the direction corresponding to a region of interest centered at the approximate position of the object of interest, are identified. These scan configurations of the LIDAR match the region of interest. The geometric characteristics of this region of interest may depend on the size and / or the shape of the object of interest. For example, if the LIDAR executes gradually a 360-degree scan pattern involving multiple horizontal scan angles and multiple vertical scan angles, the region of interest may be given by the horizontal scan angle that best matches the approximate position of the object of interest, and all the vertical scan angles. In this example, the region of interest may alternatively be given by the horizontal scan angle that best matches the approximate position of the object of interest and its direct (horizontal scan angle) neighbors, and all the vertical scan angles. Of course, the region of interest may correspond only to the approximate position of the object of interest. At substep S730, a subset of data points is selected from the obtained point cloud in a way that each data point of the selected subset is obtainable using an identified scan configuration. Preferably, the data points of the selected subset are also restricted to be within a same spatial, temporal, or spatio-temporal extent as the approximate position of the object of interest. In some embodiments compatible with a scenario of a single subset selected according to the selection process 700 of Figure 7, the spatial position of the object of interest may be determined to correspond substantially to the center of the region of interest by averaging, for example, the spatial positions represented by the data points of the selected subset. The associated time instance, on the other hand, may be determined as the timestamp of a data point belonging to the selected subset whose spatial position is the nearest to the determined spatial position of the object of interest, or may be determined as the midpoint between a first timestamp that is the earliest among all the timestamps of the selected subset and a second timestamp that is the latest among all the timestamps of the selected subset. In other embodiments also compatible with a scenario of a single subset selected according to the selection process 700 of Figure 7, only a group of data points within the selected subset is chosen to determine a spatial position and an associated time instance for the object of interest. For example, the group of data points may be chosen in a way to have a spatio-temporal extent less than a predetermined spatio-temporal extent threshold. Figure 8 is a flowchart of a method 800 for processing a point cloud representing an ITS environment that includes an object of interest. The method 800 of Figure 8 provides information about the object of interest, such as its position, speed and size. The point cloud may be composed of any number of frames of data points, and may be acquired by a LIDAR executing any type of LIDAR scan pattern. The LIDAR may be fixed or mobile, and the object of interest, whose motion is assumed to be independent of that of the LIDAR, may also be fixed or mobile. The method may be implemented at any ITS station, such as the ITS station described with reference to Figure 1. At step S810, receiving a point cloud representing, at least partially, the object of interest, the data points within the point cloud that represent the object of interest are identified. This may be achieved by applying an object detection process to the received point cloud, as described with reference to step S310. Also, use may be made of a bounding-box to enclose some or all of the identified data points. The geometric shape of the bounding-box may depend on that of the object of interest. For example, a rectangular parallelepiped may be used as a bounding-box for an object of interest represented by a vehicle, whereas a cylindrical bounding-box is more compatible with an object of interest represented by a pedestrian. The use of a predefined geometric shape for the bounding-box may result in one or more data points that do not represent the object of interest, but are enclosed within the bounding-box. In a first variant of step S810 compatible with a scenario where a bounding-box is used to enclose the identified data points, a refinement process is applied to the data points enclosed within the bounding-box in order to remove the data points that do not represent the object of interest. The removed data points may be determined as representing another object in the scanned area. The refinement process may use supplementary details that the data points may comprise, such as attenuation, frequency shift and phase shift. The refinement process may also use information provided by the object detection process, such as semantic classification of the data points. In a second variant of step S810 compatible with a scenario where the LIDAR is moving while acquiring and recording the point cloud, a pre-processing is applied to the point cloud before detecting the object of interest, in order to counteract the unwanted distortion caused by the motion of the LIDAR. This may be achieved by taken into consideration the LIDAR motion characteristics, such as its position, direction, speed and acceleration. In this way, the newly obtained point cloud can be assimilated to a point cloud acquired by a fixed LIDAR. At step S820, a spatial position and an associated time instance are determined for the object of interest based on one or multiple subsets selected from the data points that are identified as representing, at least partially, the object of interest. Each of the selected subsets may be determined in such a way to exhibit a spatial and / or temporal extent less than those of the identified data points. The way the subsets are selected, and the spatial position and the associated time instance are determined may be achieved according to one or more embodiments described with reference to the method 300 of Figure 3. At step S830, the speed of the object of interest is computed based on at least two pairs of spatial position and associated time instance. Preferably, each pair of spatial position and associated time instance is determined from a different frame of data points. Also, the computed speed of the object of interest may be associated with the latest time instance among the used time instances. When two or more speeds are computed for the object of interest for different time instances, an acceleration parameter for the object of interest may also be computed in step S830. At step S840, the shape of the object of interest as appearing in the point cloud is rectified, i.e., corrected, based on the determined position and the computed speed of the object of interest. In this way, the real shape of the object of interest is retrieved, making it possible to counteract the distortion caused by the motion of the object of interest. For example, each data point identified as representing the object of interest may be shifted along the motion direction of the object of interest by a distance that depends on its timestamp. More precisely, based on the computed speed of step S830, denoted V, the associated time instance of step S320, denoted To, and the timestamp of a data point identified as representing the object of interest, denoted Ti, this data point may be shifted in the point cloud by a distance, denoted Di, given by the following expression: Di= (To- T) *1 / . This processing may be applied to all the data points identified as representing the object of interest. In this way, the representation of the object of interest within the point cloud become undistorted. Also, the shape and the size of the object of interest become more accurate. At step S850, an ITS message comprising information about the object of interest is transmitted to other ITS users. The comprised information includes, but not limited to the spatial position, the associated time instance, the size, and the speed. For instance, the transmitted ITS message may be a Collective Perception Message (CPM). In this case, as specified in the ETSI technical report, the spatial position may refer to the ground position of the center of the front side of the bounding-box enclosing the object of interest. Figure 9 is a diagram illustrating an example of a hardware of a processing device 90 implementing a method for determining a spatial position and an associated time instance of an object of interest using a cloud of timestamped data points, according to embodiments of the invention. The processing device 90 may be coupled to a LIDAR 110 configured to generate the point cloud by scanning an environment comprising the object of interest. The processing device may also be coupled to a transmission unit 140 configured to generate and transmit messages including information generated by the processing device, such as the spatial position and the associated time instance of the object of interest. The processing device 90 may be implemented with a bus architecture linking together various circuits, including but not limited to a processor 91, a computer-readable memory 92, and multiple components. Each of the multiple components may be coupled to both the processor 91 and the computer-readable memory 92. A first component is a points identification unit 93 configured to identify, from the point cloud, data points that represent the object of interest. A second component is a subsets selection unit 94 configured to select, from the identified data points, one or more subsets of data points that, preferably, have a respective spatial, temporal or spatio-temporal extent less than that of the identified data points. A third component is a position and time determination unit 95 configured to determine, based on the selected subsets of data points, a spatial position and an associated time instance for the object of interest. A fourth component is a speed and shape computation unit 96 configured to compute a speed and a shape for the object of interest based on the determined spatial position and associated time instance. While the present invention has been described with reference to embodiments, it is to be understood that the invention is not limited to the disclosed embodiments. It will be appreciated by those skilled in the art that various changes and modification might be made without departing from the scope of the invention, as defined in the appended claims. All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and / or all of the steps of any method or process so disclosed, may be combined in any combination, except combinations where at least some of such features and / or steps are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise. Thus, unless expressly stated otherwise, each feature disclosed is one example only of a generic series of equivalent or similar features. In the claims, the word “comprising” does not exclude other elements or steps, and the indefinite article “a” or “an” does not exclude a plurality. The mere fact that different features are recited in mutually different dependent claims does not indicate that a combination of these features cannot be advantageously used. In the preceding embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored on or transmitted over, as one or more instructions or code, a computer-readable medium and executed by a hardware-based processing unit. Computer-readable media may include computer-readable storage media, which corresponds to a tangible medium such as data storage media, or communication media including any medium that facilitates transfer of a computer program from one place to another, e.g., according to a communication protocol. In this manner, computer-readable media generally may correspond to (1) tangible computer-readable storage media which is non-transitory or (2) a communication medium such as a signal or carrier wave. Data storage media may be any available media that can be accessed by one or more computers or one or more processors to retrieve instructions, code and / or data structures for implementation of the techniques described in this disclosure. A computer program product may include a computer-readable medium. By way of example, and not limitation, such computer-readable storage media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage, or other magnetic storage devices, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer. Also, any connection is properly termed a computer-readable medium. For example, if instructions are transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. It should be understood, however, that computer-readable storage media and data storage media do not include connections, carrier waves, signals, or other transient media, but are instead directed to non-transient, tangible storage media. Disk and disc, as used herein, includes compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc, where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of 5 computer-readable media.

Claims

1. A method for determining a position for an object using a cloud of timestamped points, the method comprising:o obtaining a set of points, among the cloud of points, representing, at least partially, the object;o selecting, from the set of points, at least one subset of points;o determining a spatial position and an associated time instance for the object, wherein the spatial position and the associated time instance are determined based on the at least one subset of points.

2. The method of claim 1, wherein selecting, from the set of points, at least one subset of points comprises:selecting, from the set of points, at least one subset of points, each of the at least one subset of points being associated with a spatio-temporal selection extent less than a spatio-temporal extent of the obtained set of points.

3. The method of claim 1, wherein selecting, from the set of points, at least one subset of points comprises:selecting, from the set of points, at least one subset of points, each of the at least one subset of points being associated with a spatial selection extent less than a spatial extent of the obtained set of points.

4. The method of claim 1, wherein selecting, from the set of points, at least one subset of points comprises:selecting, from the set of points, at least one subset of points, each of the at least one subset of points being associated with a temporal selection extent less than a temporal extent of the obtained set of points.

5. The method of any one of claims 2 to 4, wherein the selection extent is less than the ratio of the corresponding extent of the obtained set of points to the number of the selected subsets of points.

6. The method of any one of claims 2 to 4, wherein selecting, from the set of points, at least one subset of points further comprises:o determining a bounding-box enclosing the obtained set of points;o identifying an orientation for the object;o selecting, as a first subset of points, a first group of points that are closer to a first face of the bounding-box substantially perpendicular to the identified orientation than to a second face of the bounding-box that is opposite to the first face, andselecting, as a second subset of points, a second group of points that are closer to the second face of the bounding-boxthan to the first face.

7. The method of any one of claims 2 to 4, wherein selecting, from the set of points, at least one subset of points further comprises:selecting, as a single subset of points, a group of points that are representative of the whole object.

8. The method of any one of claims 2 to 4, wherein selecting, from the set of points, at least one subset of points further comprises:selecting, as a single subset of points, a group of points that are representative of a specific part of the object.

9. The method of claim 1, wherein the cloud of timestamped points is obtained using a LIDAR executing, at least once, a LIDAR scan pattern.

10. The method of claim 9 and any one of claims 2 to 4, wherein selecting, from the set of points, at least one subset of points further comprises:o determining an approximate position for the object;o identifying a plurality of scan configurations enabling the LIDAR to scan in the direction corresponding to a region of interest centered at the approximate position of the object;o selecting, as a single subset of points, a group of points that are obtainable using identified scan configurations.

11. The method of any one of the preceding claims, wherein determining a spatial position and an associated time instance for the object comprises, for each of the at least one subset of points:o determining an approximate spatial position for the object, based on positions associated with the points of the subset;o determining an approximate time instance for the object, based on timestamps associated with the points of the subset.

12. The method of claim 11, wherein the spatial position of the object is determined as a barycenter of the determined at least one approximate spatial position, and wherein the associated time instance is determined by averaging the determined at least one approximate time instance.

13. A method of communication in an Intelligent Transport System (ITS), comprising: transmitting an ITS message comprising items of information describing an object of interest, the items of information include a spatial position and an associated timeinstance, wherein the spatial position and the associated time instance are determined according to any one of claims 1 to 12.

14. A processing device configured to perform a method according to any one of claims 1 to 13.

515. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out a method according to any one of claims 1 to 13.10 16. A computer-readable medium carrying a computer program according to claim 15.

Images