Object positioning method, object tracking method, vehicle tracking method in road network scene, object tracking system, computing device, computer readable storage medium and computer program product

Abstract

The embodiment of the invention provides an object positioning method, an object tracking method, a vehicle tracking method in a road network scene, an object tracking system, a computing device, a computer readable storage medium and a computer program product. The object positioning method comprises the following steps: acquiring initial positioning data of a target object collected by a first positioning sensor; the initial positioning data is adjusted based on position offset information, target positioning data of the target object is obtained, the position offset information is obtained based on position matching of a plurality of sample objects in a view angle intersection area collected by all positioning sensors in a sensor group, and the sensor group comprises a first positioning sensor and a second positioning sensor; the view angle intersection area is an intersection area of view angle areas of the positioning sensors in the sensor group. According to the method, the overall offset of the positioning sensor is estimated by performing position matching on the multiple sample objects, so that the positioning data collected under the scene offset condition is adjusted, and the object positioning accuracy is improved.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of data processing technology, and in particular to an object positioning method, an object tracking method, a vehicle tracking method in a road network scenario, an object tracking system, a computing device, a computer-readable storage medium, and a computer program product. Background Art

[0002] With the development of data processing technologies such as cloud computing, big data, and artificial intelligence, object localization, as a basic task in the field of computer vision, has been widely used in many fields such as smart transportation, security monitoring, environmental monitoring, personnel search and rescue, sports broadcasting, and industrial automation.

[0003] Currently, object positioning technology is achieved by collecting object positioning data through positioning sensors. However, in actual project scenarios, when positioning sensors are subjected to collisions, temperature changes, or insecure installation, the collected positioning data may experience scene deviations, resulting in reduced accuracy of object positioning results. Summary of the Invention

[0004] In view of this, embodiments of this specification provide an object positioning method. One or more embodiments of this specification also relate to an object tracking method, a vehicle tracking method in a road network scenario, an object tracking system, an object positioning device, an object tracking device, a vehicle tracking device in a road network scenario, a computing device, a computer-readable storage medium, and a computer program product to address technical deficiencies in the prior art.

[0005] According to a first aspect of the embodiments of this specification, there is provided an object positioning method, including:

[0006] Acquire initial positioning data of the target object collected by the first positioning sensor;

[0007] Based on the position offset information, the initial positioning data is adjusted to obtain target positioning data of the target object, wherein the position offset information is obtained based on position matching of multiple sample objects in an intersection area of ​​viewing angles collected by each positioning sensor in the sensor group, the sensor group includes a first positioning sensor, and the viewing angle intersection area is an area where viewing angle areas of each positioning sensor in the sensor group intersect.

[0008] According to a second aspect of the embodiments of this specification, there is provided an object tracking method, comprising:

[0009] Obtaining initial positioning data of the target object collected by each positioning sensor in the sensor group at multiple different times;

[0010] Based on the position offset information, adjusting each initial positioning data to obtain target positioning data of the target object at multiple different moments, wherein the position offset information is obtained by positionally matching multiple sample objects within an intersecting area of ​​viewing angles collected by each positioning sensor in the sensor group, where the viewing angle intersection area is the area where the viewing angles of each positioning sensor in the sensor group intersect;

[0011] A target object trajectory of the target object is determined based on target positioning data of the target object at a plurality of different moments in time.

[0012] According to a third aspect of an embodiment of this specification, a vehicle tracking method in a road network scenario is provided, which is applied to a cloud-side device, including:

[0013] Obtaining initial positioning data of the target vehicle at multiple different times collected by each camera in a camera group of the target road scene;

[0014] Based on the position offset information, the initial positioning data are adjusted to obtain the target positioning data of the target vehicle at multiple different time points. The position offset information is obtained by position matching multiple sample vehicles within the intersection area of ​​the viewing angles captured by each camera in the camera group. The viewing angle intersection area is the area where the viewing angles of each camera in the camera group intersect.

[0015] A target vehicle trajectory of the target vehicle in the target road scene is determined based on target positioning data of the target vehicle at multiple different moments.

[0016] According to a fourth aspect of the embodiments of this specification, there is provided an object tracking system, comprising:

[0017] The object detection unit is configured to obtain the viewing area information collected by each positioning sensor in the sensor group at multiple different times; based on the viewing area information, perform object recognition within the viewing area of ​​each positioning sensor to determine the target object within the viewing area;

[0018] An object identification unit, used to obtain object identification information of a target object;

[0019] an object positioning unit, configured to obtain initial positioning data of the target object at multiple different moments, and adjust each initial positioning data based on position offset information to obtain target positioning data of the target object at multiple different moments, wherein the position offset information is obtained by position matching a plurality of sample objects within an intersecting area of ​​viewing angles collected by each positioning sensor, where the intersecting area of ​​viewing angles of each positioning sensor in the sensor group intersects;

[0020] The object tracking unit is used to fit the motion process of the target object based on the object identification information of the target object and the target positioning data at multiple different times to obtain the target object trajectory of the target object.

[0021] According to a fifth aspect of the embodiments of this specification, there is provided an object positioning device, comprising:

[0022] A first acquisition module is configured to acquire initial positioning data of the target object collected by the first positioning sensor;

[0023] The first adjustment module is configured to adjust the initial positioning data based on the position offset information to obtain target positioning data of the target object, wherein the position offset information is obtained based on position matching of multiple sample objects within a viewing angle intersection area collected by each positioning sensor in the sensor group, the sensor group includes the first positioning sensor, and the viewing angle intersection area is an area where viewing angle areas of each positioning sensor in the sensor group intersect.

[0024] According to a sixth aspect of the embodiments of this specification, there is provided an object tracking device, comprising:

[0025] A second acquisition module is configured to acquire initial positioning data of the target object collected by each positioning sensor in the sensor group at multiple different times;

[0026] a second adjustment module configured to adjust each initial positioning data based on position offset information to obtain target positioning data of the target object at multiple different moments, wherein the position offset information is obtained by position matching a plurality of sample objects within an intersecting area of ​​viewing angles collected by each positioning sensor in the sensor group, where the intersecting area of ​​viewing angles is an area where the viewing angles of each positioning sensor in the sensor group intersect;

[0027] The second tracking module is configured to determine a target object trajectory of the target object based on target positioning data of the target object at multiple different moments.

[0028] According to a seventh aspect of the embodiments of this specification, a vehicle tracking device in a road network scenario is provided, which is applied to a cloud-side device, including:

[0029] A third acquisition module is configured to acquire initial positioning data of the target vehicle at multiple different times collected by each camera in the camera group of the target road scene;

[0030] a third adjustment module configured to adjust each initial positioning data based on position offset information to obtain target positioning data of the target vehicle at multiple different time instants, wherein the position offset information is obtained by position matching multiple sample vehicles within an intersection area of ​​view angles captured by each camera in the camera group, where the intersection area of ​​view angles is an area where the view angles of each camera in the camera group intersect;

[0031] The third tracking module is configured to determine a target vehicle trajectory of the target vehicle in the target road scene based on the target positioning data of the target vehicle at multiple different time instants.

[0032] According to an eighth aspect of the embodiments of this specification, there is provided a computing device, including:

[0033] memory and processor;

[0034] The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer programs / instructions are executed by the processor, the steps of the above method are implemented.

[0035] According to a ninth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores a computer program / instruction, and the steps of the above method are implemented when the computer program / instruction is executed by a processor.

[0036] According to a tenth aspect of the embodiments of this specification, a computer program product is provided, comprising a computer program / instruction, which implements the steps of the above method when executed by a processor.

[0037] In one embodiment of the present specification, initial positioning data of a target object collected by a first positioning sensor is obtained; based on position offset information, the initial positioning data is adjusted to obtain target positioning data of the target object. The position offset information is obtained by positionally matching multiple sample objects within an intersecting area of ​​view angles collected by each positioning sensor in a sensor group, the sensor group including the first positioning sensor, and the intersecting area of ​​view angles being the area where the view angles of each positioning sensor in the sensor group intersect. Based on positionally matching multiple sample objects within the intersecting area of ​​view angles collected by each positioning sensor, overall positioning offset information of the positioning sensors is estimated, and this information is used to eliminate initial positioning data collected in the presence of scene offset, thereby improving the accuracy of the object positioning result and enhancing the object positioning effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flowchart of an object tracking method;

[0039] Figure 2 is a flow chart of an object positioning method provided by one embodiment of this specification;

[0040] Figure 3 is a schematic diagram of a positioning result before offset in an object positioning method provided by one embodiment of this specification;

[0041] Figure 4 is a schematic diagram of a positioning result after shifting in an object positioning method provided by one embodiment of this specification;

[0042] Figure 5 This is a schematic diagram of adjusting a positioning result after an offset in an object positioning method provided by one embodiment of this specification;

[0043] Figure 6 This is a schematic diagram of determining sample matching relationships in an object positioning method provided by an embodiment of this specification;

[0044] Figure 7 is a schematic diagram of local coordinate system transformation in an object positioning method provided by one embodiment of this specification;

[0045] Figure 8 is a flowchart of an object tracking method provided by one embodiment of this specification;

[0046] Figure 9 is a flowchart of an object tracking method provided by one embodiment of this specification;

[0047] Figure 10 This is a flow chart of a vehicle tracking method in a road network scenario provided by one embodiment of this specification;

[0048] Figure 11 This is a flowchart of a process for a vehicle tracking method applied to a global road network scenario provided by an embodiment of this specification;

[0049] Figure 12 This is a front-end schematic diagram of a vehicle tracking method applied to a global road network scenario provided by an embodiment of this specification;

[0050] Figure 13 is a schematic structural diagram of an object tracking system provided by one embodiment of this specification;

[0051] Figure 14 This is a schematic structural diagram of an object positioning device provided by one embodiment of this specification;

[0052] Figure 15 is a schematic structural diagram of an object tracking device provided by one embodiment of this specification;

[0053] Figure 16 This is a schematic diagram of the structure of a vehicle tracking device in a road network scenario provided by one embodiment of this specification;

[0054] Figure 17 This is a structural block diagram of a computing device provided by one embodiment of this specification. DETAILED DESCRIPTION

[0055] The following description sets forth many specific details to facilitate a thorough understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the scope of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0056] The terms used in one or more embodiments of this specification are for the purpose of describing specific embodiments only and are not intended to limit one or more embodiments of this specification. The singular forms "a," "the," and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0057] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0058] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0059] First, the terms involved in one or more embodiments of this specification are explained.

[0060] Computer vision is the science that studies how machines can "see." Specifically, it refers to the use of visual sensors and computers to replace the human eye in identifying, tracking, and measuring objects. This is followed by image processing, where computers create images more suitable for human observation or for transmission to instruments. As a scientific discipline, computer vision studies related theories and technologies, aiming to build artificial intelligence systems capable of extracting information from images or multidimensional data.

[0061] Object Detection: It aims to identify which target objects exist in an image or video frame and determine their locations in the image or video frame.

[0062] Object Identification: Object identification can be considered as a part of object detection, especially when it involves identifying and distinguishing different object instances of the same category. It refers to the ability to uniquely identify a specific object through unique features or markers, for example, distinguishing two similar vehicles in a multi-target tracking scenario.

[0063] Object Localization: Object localization is the process of determining the exact location of a specific object in an image. This is usually expressed as providing a bounding box or a finer pixel-level outline surrounding the object. It is one of the basic steps of object detection and focuses on the question of "where" but not "what".

[0064] Object Tracking: Object tracking is the process of continuously and stably following and locating the same target object in a continuous image sequence or video stream. Building on object detection, it predicts and updates the target object's position in the current frame based on previous state, maintaining tracking even if the target object moves, is obscured, or experiences changes in lighting. This process requires algorithms to adapt to dynamic changes in the target, ensuring consistent tracking across time and space.

[0065] Latitude and longitude: Latitude and longitude are the combined names of longitude and latitude, forming a coordinate system called the geographic coordinate system. It is a spherical coordinate system that uses a three-dimensional sphere to define space on the earth and can mark any location on the earth.

[0066] The SL coordinate system (SL Coordinate System) is a curvilinear coordinate system commonly used in autonomous driving, road modeling, and vehicle path planning. In this coordinate system, the position of a point relative to the road centerline is described by defining two orthogonal direction quantities.

[0067] S-axis (Longitudinal Axis or S-coordinate): The direction extending along the centerline of the road, usually used to represent the longitudinal position of an object along the road, that is, the distance along the direction of the road.

[0068] L-axis (LateralAxis or L-coordinate): The direction perpendicular to the S-axis and pointing to the inside of the road. It is used to represent the lateral offset distance of an object relative to the centerline of the road.

[0069] Hungarian matching: A combinatorial optimization algorithm for solving task allocation problems in polynomial time.

[0070] KM (Kuhn-Munkres) algorithm: an optimization algorithm for solving the maximum weight matching problem in bipartite graphs.

[0071] Greedy Algorithm: An optimization algorithm that adopts the current state of the target decision (local target solution) at each step, attempting to gradually build a global target solution. At each step, the algorithm always makes the choice that appears to be optimal at the moment, without considering possible future consequences. In positioning systems, greedy algorithms may be used to quickly obtain preliminary matching results.

[0072] Kalman Filter: Kalman filter is a commonly used algorithm for estimating the state of a dynamic system, especially suitable for processing noisy data.

[0073] Hidden Markov Model (HMM): A statistical modeling tool that describes a random process with hidden states. In this model, the true state of the system is not directly observable (i.e., "invisible"), but the series of observable output signals or events generated by these states can be observed. Each hidden state transitions to the next state with a certain probability, and each state generates a set of specific probability distributions that determine the observable data generated at the next moment.

[0074] At present, object tracking supported by object positioning has gradually shifted from a single positioning sensor to multiple positioning sensors, realizing full-area tracking of the target object. Figure 1 A flow chart of an object tracking method is shown in FIG. Figure 1 As shown:

[0075] First, object detection: obtain the viewing area information collected by each positioning sensor in the sensor group at multiple different times; based on the viewing area information, perform object recognition in the viewing area of ​​each positioning sensor to determine the target object in the viewing area.

[0076] Next, object identification: obtain the object identification information of the target object.

[0077] Then, object positioning: obtain the positioning data of the target object at multiple different times.

[0078] Finally, object tracking, based on the positioning data at multiple different times, as well as the target box position, target box size, target longitude and latitude, and object identification (object ID), fits the motion process of the target object to obtain the target object trajectory of the target object.

[0079] However, when the positioning sensor is subjected to collision, temperature changes, or is not installed firmly, the collected positioning data will be subject to scene deviation, resulting in reduced accuracy of the object positioning results, and then causing the overall deviation of the target object trajectory obtained by object tracking, reducing the effectiveness of object tracking.

[0080] In this specification, an object positioning method is provided. This specification also relates to an object tracking method, a vehicle tracking method in a road network scenario, an object tracking system, an object positioning device, an object tracking device, a vehicle tracking device in a road network scenario, a computing device, a computer-readable storage medium and a computer program product, which are described in detail one by one in the following embodiments.

[0081] See also Figure 2 , Figure 2 A flowchart of an object positioning method provided by an embodiment of this specification is shown, including the following specific steps:

[0082] Step 202: Acquire initial positioning data of the target object collected by a first positioning sensor.

[0083] The embodiments of this specification are applicable to intelligent system platforms with object positioning capabilities, such as smart transportation systems, security monitoring systems, environmental monitoring systems, sports broadcasting platforms, automated warehouse management platforms, robotic navigation systems, unmanned vehicle positioning platforms, or industrial production systems based on Internet of Things technology. In these system platforms, positioning data collected by positioning sensors is used to accurately locate objects and thus implement the corresponding functions.

[0084] A positioning sensor is a device used to collect positioning data for objects in a scene. The positioning sensor has a corresponding viewing area and collects positioning data for objects within the viewing area. Positioning sensors include, but are not limited to, cameras, infrared sensors, radar sensors, and radio frequency sensors. Examples include road monitoring cameras installed in road scenes, infrared sensors installed in security monitoring scenes, radar sensors installed in unmanned vehicles, and radio frequency sensors installed on production lines. The first positioning sensor is the positioning sensor in the sensor group used to collect positioning data for the target object.

[0085] Target objects are entities to be located, including but not limited to people, vehicles, robots, environmental objects, and goods. Examples include pedestrians and vehicles in road scenarios, people in security surveillance scenarios, vehicles, pedestrians, and environmental objects in autonomous driving scenarios, and workers and goods on a production line.

[0086] Positioning data is data information collected by positioning sensors that describes the position, posture, and motion state (e.g., direction, speed, and acceleration) of an object in a specific spatial coordinate system, including but not limited to: longitude and latitude, three-dimensional spatial coordinates, distance, angle, speed, acceleration, angular velocity, angular acceleration, direction, and trajectory (road). For example, the positioning data of a vehicle in a road scenario is the vehicle's longitude and latitude, direction, and road. Another example is the three-dimensional spatial coordinates, distance, and angle of a person in a security monitoring scenario. Another example is the robot's positioning data in a robot navigation scenario, which is the robot's distance, speed, acceleration, angular velocity, angular acceleration, direction, and trajectory. Another example is the distance, speed, and trajectory of goods products on a production line.

[0087] The initial positioning data of the target object is the positioning data of the target object directly collected by the first positioning sensor. Since the first positioning sensor may be affected by collisions, temperature changes, or loose installation, the accuracy of the initial positioning data cannot be guaranteed. The initial positioning data serves as the basic data for subsequent optimization processing.

[0088] It should be noted that when the first positioning sensor is installed, reference positioning data is preset for the first positioning sensor. The first positioning sensor determines the initial positioning data of the target object based on this reference positioning data. For example, in a security monitoring scenario, an infrared sensor is pre-installed on the ceiling. First, reference positioning data needs to be preset for it: its fixed coordinates in three-dimensional space (X0, Y0, Z0) and orientation angle θ are set. When a person enters the monitoring area, the infrared sensor captures the person's heat signal and uses the person's position coordinates (px, py, pz) within the viewing area. This initial position coordinate is relative to the camera's own coordinate system. Based on this reference positioning data, a coordinate conversion is performed to obtain the person's initial three-dimensional space coordinates (X0+px, Y0+py, Z0+pz) = (X, Y, Z).

[0089] Optionally, before step 202, the following specific steps are further included:

[0090] Obtaining viewing area information collected by the first positioning sensor;

[0091] Based on the viewing angle area information, object recognition is performed within the viewing angle area of ​​the first positioning sensor to obtain a target object within the viewing angle area and initial positioning data of the target object.

[0092] When the positioning sensor is a camera, the captured viewing area information is an image or video. Object recognition is achieved by performing entity detection on the image or video frame using a target frame. Specific entity detection can be performed using artificial intelligence algorithms, template matching, visual feature point analysis, etc.

[0093] For example, in a smart transportation system, multiple road monitoring cameras are pre-installed in a road scene, and four vehicles are traveling on a single three-lane road on Road M. The road monitoring camera on Road M captures the viewing area in which it is located and uploads the road image. Entity detection is performed on the road image using a target frame, and four vehicles are identified as traveling in the viewing area. Based on the baseline longitude and latitude [longitude_0, latitude_0] of the road monitoring camera, the initial longitude and latitude of the four vehicles and their lanes are determined: vehicle 1 {[longitude_1, latitude_1], lane 3}, vehicle 2 {[longitude_2, latitude_2], lane 2}, vehicle 3 {[longitude_3, latitude_3], lane 1}, and vehicle 4 {[longitude_4, latitude_4], lane 3}.

[0094] Acquiring the initial positioning data of the target object collected by the first positioning sensor provides basic data support for subsequent positioning data adjustment and object positioning.

[0095] Step 204: Based on the position offset information, adjust the initial positioning data to obtain target positioning data of the target object, wherein the position offset information is obtained based on position matching of multiple sample objects within a viewing angle intersection area collected by each positioning sensor in the sensor group, the sensor group includes a first positioning sensor, and the viewing angle intersection area is an area where the viewing angle areas of each positioning sensor in the sensor group intersect.

[0096] The displacement offset information is the offset information of the object position estimated for the scene deviation caused by the positioning sensor. The position offset information may be directional or non-directional, which is not limited here. The position offset information may be a specific quantitative value or an offset matrix (vector) used to adjust the positioning data of the artificial intelligence model, which is not limited here. The position offset information is obtained based on the position matching of multiple sample objects in the intersection area of ​​the viewing angle collected by each positioning sensor in the sensor group, and characterizes the difference between the positioning data obtained by the scene offset and the real position. The position offset information can be the position offset information of multiple sample objects, or it can be an overall target position offset information, which is not limited here. For example, in a smart transportation system, two adjacent road monitoring cameras simultaneously monitor the same vehicle. One of the road monitoring cameras was hit by a vehicle before and had a position deviation. There is a deviation from the basic positioning data set in advance for the road monitoring camera. The longitude and latitude of the vehicle collected by the two road monitoring cameras are quite different, and it is impossible to determine which set of longitude and latitude is the real longitude and latitude of the vehicle. Based on the position matching of multiple sample vehicles in the intersection area of ​​the viewing angles captured by each positioning road monitoring camera in the camera group, the latitude and longitude offset (+0.005, -0.002) is estimated.

[0097] A sensor group is a device combination formed by combining multiple positioning sensors. Each of the multiple positioning sensors collects positioning data for an object. Based on these multiple sets of positioning data, global tracking of the object and estimation of offset information can be achieved, enabling more comprehensive and more accurate object positioning. The viewing angles of each positioning sensor in the sensor group may intersect. For example, at an intersection, four road monitoring cameras are installed on four directional roads. These four road monitoring cameras are combined into a sensor group. Sensor groups can be pre-set, for example, on a generated line, two adjacent radio frequency sensors are determined to belong to one sensor group. Sensor groups can also be flexibly set according to positioning requirements. For example, in a security monitoring scenario, at least one infrared sensor with an intersecting viewing angle is determined to belong to one sensor group.

[0098] The target positioning data of the target object is the positioning result of the target object after adjustment and optimization. Compared with the initial positioning data, the target positioning data is a more accurate positioning data and is the object positioning result.

[0099] The view intersection area is the area where the view angles of each positioning sensor in a sensor group intersect. Objects within the view intersection area are recognized by all positioning sensors and their positioning data is collected. This area is used to cross-validate and adjust the positioning data collected by any positioning sensor in the sensor group. For example, four road monitoring cameras installed on the four sides of an intersection each have a view angle for that intersection. This intersection is the view intersection area of ​​these four road monitoring cameras.

[0100] The sample object is a pre-located object entity. The sample object is recognized by each positioning sensor in the sensor group within the intersection area of ​​the viewing angles, and sample positioning data is collected. The time point at which the sample object is collected by each positioning sensor in the sensor group precedes the time point of step 202.

[0101] Position matching is the process of determining offset information based on the matching relationship between the sample positioning data of the sample objects and the sample objects within the view intersection region. Optionally, based on the matching relationship between the sample positioning data of the sample objects and the sample objects within the view intersection region, position offset information of multiple sample objects within the view intersection region captured by each positioning sensor is determined, and target position offset information is determined based on the position offset information. For example, the view intersection region contains three sample objects. The three sample objects within the view intersection region captured by the first positioning sensor are sample objects A, sample objects B, and sample objects C, and the three sample objects within the view intersection region captured by the second positioning sensor are sample objects D, sample objects E, and sample objects F. The matching relationship is determined as follows: sample objects A and sample objects D are the same sample object, sample objects B and sample objects E are the same sample object, and sample objects C and sample objects F are the same sample object. Based on the first sample positioning data of the three sample objects captured by the first positioning sensor and the second sample positioning data of the three sample objects captured by the second positioning sensor, position offset information of the three sample objects is determined. Based on the three position offset information, target position offset information is determined. For details, see the subsequent description.

[0102] When the position offset information is a specific quantized value, the initial positioning data is adjusted based on the position offset information to obtain the target positioning data of the target object. Specifically, the position offset information is added (subtracted) from the initial positioning data to obtain the target positioning data of the target object.

[0103] When the position offset information is an offset matrix (vector), the initial positioning data is adjusted based on the position offset information to obtain the target positioning data of the target object. The specific method is: using an artificial intelligence model, based on the position offset information, the initial positioning data is adjusted to obtain the target positioning data of the target object.

[0104] For example, based on the sample longitudes and latitudes and lanes of 20 sample vehicles (vehicle_i, i to [1, 20]), and the matching relationships between the 20 sample vehicles within the intersection region of the viewing angles captured by the road monitoring cameras in the M-road and N-road camera groups, 20 position offsets of the 20 sample vehicles within the intersection region of the viewing angles are determined: [longitude offset_i, latitude offset_i, i to [1, 20]]. Based on these 20 position offsets, the average value is determined as the target position offset [longitude offset, latitude offset]. Add the target position offset [longitude offset, latitude offset] to the initial longitude and latitude of the four vehicles to obtain the target longitude and longitude and lanes of the four vehicles: vehicle 1 {[longitude_1+longitude offset, latitude_1+latitude offset], lane 3}, vehicle 2 {[longitude_2+longitude offset, latitude_2+latitude offset], lane 2}, vehicle 3 {[longitude_3+longitude offset, latitude_3+latitude offset], lane 1} and vehicle 4 {[longitude_4+longitude offset, latitude_4latitude offset], lane 3}.

[0105] For the specific implementation process of step 202 and step 204, see Figure 3 、 Figure 4 and Figure 5 .

[0106] Figure 3 A schematic diagram of a positioning result before offset in an object positioning method provided by an embodiment of this specification is shown, as shown in the figure:

[0107] The sensor group includes a first positioning sensor and a second positioning sensor. The intersection of the viewing angles captured by the two positioning sensors contains four target objects: target object 1, target object 2, target object 3, and target object 4. Before scene drift occurs, the deviation between the positioning data of the four target objects collected by the first positioning sensor and the positioning data of the four target objects collected by the second positioning sensor is small.

[0108] Figure 4 FIG. 1 shows a schematic diagram of a positioning result after shifting in an object positioning method provided by an embodiment of the present specification, such as Figure 4 As shown:

[0109] The sensor group includes a first positioning sensor and a second positioning sensor. The intersection of the perspectives captured by the two positioning sensors contains four target objects: target object 1, target object 2, target object 3, and target object 4. After scene shift occurs, the deviation between the positioning data of the four target objects collected by the first positioning sensor and the positioning data of the four target objects collected by the second positioning sensor increases.

[0110] After the scene shift occurs, although the deviation becomes larger, because it is an overall shift, the relative positions of the positioning data of the four target objects collected by the same positioning sensor remain basically unchanged. Therefore, the methods of steps 202 and 204 above can be used to adjust the positioning data as a whole.

[0111] Figure 5 FIG. 1 shows a schematic diagram of adjusting the positioning result after offset in an object positioning method provided by an embodiment of this specification, such as Figure 5 As shown:

[0112] The sensor group includes a first positioning sensor and a second positioning sensor. The intersection of the perspectives captured by the two positioning sensors contains four target objects: target object 1, target object 2, target object 3, and target object 4. After scene offset occurs, the positioning data of the four target objects collected by the second positioning sensor is adjusted based on the position offset information. At this time, the deviation between the positioning data of the four target objects collected by the first positioning sensor and the positioning data of the four target objects collected by the second positioning sensor is reduced.

[0113] In the embodiments of this specification, based on position matching of multiple sample objects in the intersection area of ​​the perspectives collected by each positioning sensor, the overall positioning offset information of the positioning sensor is estimated, and this is used to eliminate the initial positioning data collected under the scene offset condition, thereby improving the accuracy of the object positioning result and improving the object positioning effect.

[0114] In an optional embodiment of this specification, before step 204, the following specific steps are further included:

[0115] Acquire sample positioning data of multiple sample objects in the view intersection area collected by each positioning sensor in the sensor group, as well as initial matching relationships of the multiple sample objects in the view intersection area;

[0116] Based on the initial matching relationship, position matching is performed on multiple sample objects to obtain sample matching relationships of the multiple sample objects;

[0117] Position offset information is determined based on the sample positioning data and the sample matching relationship of the plurality of sample objects.

[0118] Generally, within a certain period of time before step 204, based on position matching of multiple sample objects within the intersection area of ​​the viewing angles collected by each positioning sensor, the overall positioning offset information of the positioning sensor is estimated, and the initial positioning data with insufficient accuracy can be adjusted.

[0119] The sample positioning data of the sample object is the positioning data of the sample object directly collected by each positioning sensor in the sensor group. Since it is collected within a certain period of time, the sample positioning data reflects the regularity of scene offset to a certain extent. That is, similar to the initial positioning data, it may also be because the positioning sensor may be affected by collision, temperature changes, or loose installation. The accuracy of the sample positioning data cannot be guaranteed. The sample positioning data serves as reference data for subsequent object positioning.

[0120] The initial matching relationship of the sample objects in the intersection area of ​​the perspective is the object matching relationship of the sample objects in the intersection area of ​​the perspectives collected by each positioning sensor. If the sample objects in the intersection area of ​​the perspectives collected by each positioning sensor are the same sample objects, then they have an initial matching relationship. Due to scene offset or performance constraints of the positioning sensor, the initial matching relationship may be inaccurate. There may be a situation where the sample object in the intersection area of ​​the perspective collected by any positioning sensor corresponds to multiple sample objects in the intersection area of ​​the perspectives collected by other positioning sensors. For example, in a warehousing scenario, three RF sensors are installed in adjacent shelf aisles, and these three RF sensors form a sensor group. During a certain period of time, three sample robots A, B, and C passed through the intersection area of ​​the perspectives of the three RF sensors in a front-to-back position. While three sample robots are detected in the intersection area of ​​the viewing angle of the first RF sensor, three sample robots are also detected in the intersection area of ​​the viewing angles of the second and third RF sensors. Due to possible scene offset or RF sensor performance, the first RF sensor determines the front and back positions of the robot to be A, B, and C, the second RF sensor determines the front and back positions of the robot to be B, A, and C, and the third RF sensor determines the front and back positions of the robot to be A, B, and C. The initial matching relationship is: sample robot A - sample robot B, sample robot A, sample robot B - sample robot A, sample robot B, sample robot C - sample robot C.

[0121] The sample matching relationship for a sample object is the object matching relationship within the intersecting area of ​​the perspectives captured by each positioning sensor, obtained after adjustment and optimization. For this sample matching relationship, a sample object within the intersecting area of ​​the perspectives captured by any positioning sensor corresponds to only one sample object within the intersecting area of ​​the perspectives captured by any other positioning sensor. Compared to the initial matching relationship, the sample matching relationship is a matching relationship with higher accuracy. For example, based on the three initial matching relationships described above, position matching is performed on three sample robots to obtain the sample matching relationships for the three sample objects: sample robot A - sample robot A, sample robot B - sample robot B, and sample robot C - sample robot C.

[0122] Optionally, obtaining sample positioning data of a plurality of sample objects in the view intersection region collected by each positioning sensor in the sensor group, as well as initial matching relationships of the plurality of sample objects in the view intersection region, includes the following specific steps:

[0123] According to a preset frequency, sample positioning data of multiple sample objects in the viewing angle intersection area collected by each positioning sensor in the sensor group and initial matching relationships of the multiple sample objects in the viewing angle intersection area are obtained.

[0124] Since it is uncertain when and which positioning sensor will experience scene offset, the more timely the position offset information is determined, the higher the accuracy of the position offset information. For example, when the temperature difference between morning and evening is large, the positioning sensor will experience scene offset. Sample positioning data and initial matching relationships can be updated every hour to determine the corresponding position offset information. Alternatively, sample positioning data and initial matching relationships can be obtained every 10 minutes to determine the corresponding position offset information, and the average value for one hour can be taken.

[0125] Based on the initial matching relationship, position matching is performed on the multiple sample objects to obtain sample matching relationships for the multiple sample objects. Specifically, under a constraint on the number of matching relationships, position matching is performed on the multiple sample objects based on the initial matching relationship to obtain sample matching relationships for the multiple sample objects. The constraint on the number of matching relationships includes, but is not limited to, Hungarian matching, KM matching, and greedy algorithms.

[0126] Determine position offset information based on the sample positioning data and sample matching relationships of multiple sample objects. Specifically, obtain position offset information through data statistics based on the positioning data of sample objects with sample matching relationships. The statistical methods may include calculating the arithmetic mean, weighted mean, least squares method, or median filtering.

[0127] Figure 6 FIG. 1 shows a schematic diagram of determining a sample matching relationship in an object positioning method provided by an embodiment of this specification, such as Figure 6 As shown:

[0128] The multiple first sample objects within the intersecting area of ​​the viewing angles captured by the first positioning sensor include: sample object A, sample object B, sample object C, and sample object D. The multiple second sample objects within the intersecting area of ​​the viewing angles captured by the second positioning sensor include: sample object E, sample object F, and sample object G. Sample object A has an initial matching relationship with sample objects E and sample object F, sample object B has an initial matching relationship with sample objects F and sample object G, sample object C has an initial matching relationship with sample objects E and sample object F, and sample object D has an initial matching relationship with sample object F.

[0129] After position matching, it is determined that sample object A and sample object F have a sample matching relationship, sample object B and sample object G have a sample matching relationship, sample object C and sample object E have a sample matching relationship, and sample object D does not have a sample matching relationship.

[0130] For example, every 10 minutes, the sample longitude and latitude of 20 sample vehicles (vehicle_i, i~[1,20]) within the view intersection area captured by each road monitoring camera in the M and B road camera groups are obtained, as well as the initial matching relationship of the 20 sample vehicles within the view intersection area. Using the Hungarian matching method, based on the initial matching relationship, the 20 sample vehicles are positionally matched to obtain a sample matching relationship for the 20 sample vehicles. The arithmetic average of the sample longitude and latitude of the sample vehicles with a sample matching relationship is calculated to determine the average position offset of the 20 sample vehicles within the view intersection area: [E(longitude offset_i), E(latitude offset_i), i~[1,20]].

[0131] In the embodiments of this specification, position matching is performed on multiple sample objects based on the initial matching relationship to obtain a sample matching relationship for the multiple sample objects, thereby ensuring that the position offset information is determined based on the sample matching relationship, which has a higher accuracy, thereby improving the accuracy of the position offset information and further improving the accuracy of subsequent object positioning results.

[0132] In an optional embodiment of the present specification, position matching is performed on multiple sample objects based on the initial matching relationship to obtain a sample matching relationship of the multiple sample objects, including the following specific steps:

[0133] Based on the initial matching relationship, position matching is performed on multiple sample objects to obtain target matching relationships of the multiple sample objects;

[0134] When the number of target matching relationships of the multiple sample objects reaches a preset threshold, the target matching relationships of the multiple sample objects are determined to be sample matching relationships of the multiple sample objects.

[0135] Since the sample matching relationship is a strict one-to-one correspondence, it is necessary to determine the sample matching relationship from the perspective of global matching to avoid a large number of sample objects without a determined matching relationship. Figure 6 For example, if only sample object D does not have a sample matching relationship among the determined sample matching relationships, and if the determined sample matching relationships are as follows: sample objects A and sample objects E have a sample matching relationship, and sample objects B and sample objects F have a sample matching relationship, but in this case, sample objects C and sample objects D do not have a sample matching relationship. Therefore, it is necessary to constrain the number of matching relationships.

[0136] The target matching relationship for a sample object is the reference object matching relationship for the adjusted and optimized sample object within the intersecting area of ​​the perspectives captured by each positioning sensor. This target matching relationship is not necessarily the final sample matching relationship. For this target matching relationship, a sample object within the intersecting area of ​​the perspectives captured by any positioning sensor corresponds to only one sample object within the intersecting areas of the perspectives captured by any other positioning sensor. In the embodiments of this specification, position matching is a dynamic process, and the target matching relationship is the reference result of this dynamic process. A target matching relationship is determined as a sample matching relationship only when the number of target matching relationships reaches a preset threshold.

[0137] The number of target matching relationships is the number of target matching relationships (reference object matching relationships) determined during the position matching process for multiple sample objects. This number serves as a metric for evaluating matching results. For example, if 20 sample objects are matched within the intersecting area of ​​view captured by multiple positioning sensors, and if, after position matching, 18 pairs of sample objects are determined to have target matching relationships, the number of target matching relationships is 18.

[0138] The preset threshold is a pre-set judgment condition for the number of target matching relationships, and is a judgment condition for evaluating matching results, which is used to achieve the best possible global matching. The preset threshold can be a specific numerical value, or it can be a ratio (not greater than 1) between the number of target matching relationships and the number of all sample objects, and is not limited here. In the embodiments of this specification, the preset threshold is used to measure whether the number of target matching relationships is large enough to ensure that the obtained sample matching relationships represent the global sample objects.

[0139] Based on the initial matching relationship, position matching is performed on multiple sample objects to obtain a target matching relationship of the multiple sample objects. Specifically, based on the initial matching relationship, the matching relationship of the multiple sample objects in the intersection area of ​​the perspectives collected by each positioning sensor is determined to obtain a target matching relationship of the multiple sample objects.

[0140] by Figure 6 As an example, the embodiments of this specification are described:

[0141] The first position matching determines the target matching relationship between two sample objects: sample object A and sample object E have a target matching relationship, and sample object B and sample object F have a target matching relationship. The number of target matching relationships is 2, and the preset threshold is 3, which does not reach the preset threshold.

[0142] The second position matching determines target matching relationships between three sample objects: sample object A and sample object F have a target matching relationship, sample object B and sample object G have a target matching relationship, and sample object C and sample object E have a target matching relationship. The number of target matching relationships is 3, and the preset threshold is 3, which is reached.

[0143] Exemplarily, based on the initial matching relationship of 20 sample vehicles (vehicle_i, i~[1,20]) in the perspective intersection area, the matching relationship of multiple 20 sample vehicles in the perspective intersection area captured by each road monitoring camera is determined, and the target matching relationship of 18 sample vehicles is obtained. When the number of target matching relationships (18) reaches a preset threshold (18), the target matching relationship of the 18 sample vehicles is determined to be the sample matching relationship of the 18 sample vehicles.

[0144] In the embodiments of this specification, the number of target matching relationships is used as a judgment indicator to achieve global matching of sample objects under quantity constraints, further improving the accuracy of sample matching relationships and further improving the accuracy of subsequent determination of position offset information.

[0145] In an optional embodiment of the present specification, the plurality of sample objects include a plurality of first sample objects within an intersecting area of ​​viewing angles captured by the first positioning sensor and a plurality of second sample objects within an intersecting area of ​​viewing angles captured by the second positioning sensor, an initial matching relationship exists between any first sample object and at least one second sample object, and the second positioning sensor is any one of the sensor group except the first positioning sensor;

[0146] Correspondingly, based on the initial matching relationship, position matching is performed on multiple sample objects to obtain a target matching relationship of the multiple sample objects, including the following specific steps:

[0147] Randomly extract a first sample extraction object from multiple first sample objects;

[0148] Randomly extract a second sample extraction object from at least one second sample extraction object that has an initial matching relationship with the first sample extraction object, and determine the initial matching relationship between the first sample extraction object and the second sample extraction object as a target matching relationship;

[0149] The first sample extraction object is deleted from the plurality of first sample objects, and the step of randomly extracting a first sample extraction object from the plurality of first sample objects is returned to be executed until a target matching relationship between each first sample object and the second sample object is obtained.

[0150] The first sample objects are sample objects within the intersecting area of ​​the viewing angles collected by the first positioning sensor. The second sample objects are sample objects within the intersecting area of ​​the viewing angles collected by the second positioning sensor.

[0151] There is an initial matching relationship between any first sample object and at least one second sample object. Figure 6 .

[0152] The first sample extraction object is an object instance randomly selected from multiple first sample objects. For example, in an intelligent transportation system, to start a matching process, a vehicle is randomly selected from the sample vehicles collected by the first positioning sensor as the first sample extraction object.

[0153] The second sample object is an object instance randomly selected from multiple second sample objects that have an initial matching relationship with the first sample object. For example, if a vehicle has been selected as the first sample object, the corresponding second sample object is the same vehicle captured by the second positioning sensor, determined based on the initial matching relationship between the two.

[0154] by Figure 6 As an example, the embodiments of this specification are described:

[0155] First position matching: Randomly extract one first sample extraction object, sample object A, from the four first sample objects (sample object A, sample object B, sample object C, and sample object D). Randomly extract one second sample extraction object, sample object E, from the two second sample objects (sample object E and sample object F) that have an initial matching relationship with sample object A. The initial matching relationship between sample object A and sample object E is determined as the target matching relationship. Sample object A is removed from the four first sample objects, and a random first sample extraction object, sample object B, is extracted from the three first sample objects (sample object B, sample object C, and sample object D). This process continues until a target matching relationship is obtained between each first sample object and the second sample object: sample object A and sample object E have a target matching relationship, and sample object B and sample object F have a target matching relationship. At this point, the number of target matching relationships is 2, and the preset threshold is 3, which does not meet the preset threshold.

[0156] Second position matching: Randomly extract a first sample extraction object, sample object A, from the four first sample objects (sample object A, sample object B, sample object C, and sample object D). Randomly extract a second sample extraction object, sample object F, from the two second sample objects (sample object E and sample object F) that have an initial matching relationship with sample object A. The initial matching relationship between sample object A and sample object F is determined as the target matching relationship. Remove sample object A from the four first sample objects and randomly extract a first sample extraction object, sample object B, from the three first sample objects (sample object B, sample object C, and sample object D). This process continues until a target matching relationship is obtained between each first sample object and the second sample object: sample object A has a target matching relationship with sample object F, sample object B has a target matching relationship with sample object G, and sample object C has a target matching relationship with sample object E. At this point, the number of target matching relationships is 3, and the preset threshold is 3, which has been reached.

[0157] For example, a first sample vehicle (vehicle_j, j to [1, 20]) is randomly selected from the 20 first sample vehicles. A second sample vehicle (vehicle_k) is randomly selected from at least one second sample vehicle that has an initial matching relationship with the first sample vehicle, and the initial matching relationship between the first sample vehicle and the second sample vehicle is determined to be the target matching relationship. The first sample vehicle (vehicle_j) is deleted from the 20 first sample vehicles, and the process returns to randomly selecting a first sample vehicle from the 19 first sample vehicles (vehicle_j, j to [1, 19]), until the target matching relationship between 18 first sample vehicles and second sample vehicles is obtained.

[0158] In the embodiments of this specification, the Hungarian matching method is used to determine the sample matching relationship, and then determine the position offset information, further improving the accuracy of the position offset information, thereby eliminating the initial positioning data collected under the scene offset condition, improving the accuracy of the object positioning result, and improving the object positioning effect. In addition, the principle and process are simple and the implementation cost is low.

[0159] In an optional embodiment of the present specification, determining position offset information based on sample positioning data and sample matching relationships of multiple sample objects includes the following specific steps:

[0160] Calculating offset information of sample positioning data of the multiple sample objects based on the sample matching relationship of the multiple sample objects;

[0161] Based on the offset information of each sample positioning data, position offset information is determined.

[0162] The offset information of the sample positioning data of multiple sample objects is the offset information between the sample positioning data of sample objects with a sample matching relationship, which measures the difference between the positioning data obtained by scene offset. The offset information can be directional or non-directional, which is not limited here. The offset information can be a specific quantitative value or an offset matrix (vector) used to adjust the positioning data of the artificial intelligence model, which is not limited here. For example, in a certain perspective intersection area, multiple sample vehicles are jointly identified by two road monitoring cameras. Based on the sample matching relationship, it is determined that a certain sample vehicle in the perspective intersection area captured by the two road monitoring cameras is sample vehicle A. The positioning data of sample vehicle A under the first road monitoring camera is [longitude_A1, latitude_A1]; the positioning data of sample vehicle A under the second road monitoring camera is [longitude_A2, latitude_A2]; the offset information of the sample positioning data is [longitude offset information_A = longitude_A2 - longitude_A1, latitude offset information_A = latitude_A2 - latitude_A1].

[0163] Based on the sample matching relationship of multiple sample objects, the offset information of the sample positioning data of the multiple sample objects is calculated. The specific method is: based on the sample matching relationship of the multiple sample objects, the sample positioning data of the multiple sample objects with the sample matching relationship are determined, and the offset information of the sample positioning data of the multiple sample objects is calculated.

[0164] Based on the offset information of each sample positioning data, the position offset information is determined by statistically analyzing the offset information of each sample positioning data to obtain the position offset information. The statistical method may be arithmetic mean, weighted mean, least squares method, or median filtering method.

[0165] Exemplarily, based on the sample matching relationship of 20 sample objects, the sample longitude and latitude of the 20 sample objects with a sample matching relationship are determined, and the offsets of the sample longitude and latitude of the 20 sample objects are calculated. The arithmetic average of the offsets of the 20 sample positioning data is calculated to obtain the average position offset: [E(longitude offset_i), E(latitude offset_i), i~[1, 20]].

[0166] In the embodiments of this specification, position offset information is effectively estimated, the accuracy of the position offset information is further improved, and the accuracy of subsequent object positioning results is further improved.

[0167] In an optional embodiment of the present specification, obtaining sample positioning data of multiple sample objects in a viewing angle intersection region collected by each positioning sensor in a sensor group, as well as initial matching relationships between the multiple sample objects in the viewing angle intersection region, includes the following specific steps:

[0168] Acquire viewing area information collected by each positioning sensor in the sensor group, wherein the viewing area information is scene information within the viewing area collected by each positioning sensor;

[0169] Based on the viewing angle area information, object recognition is performed within the viewing angle area of ​​each positioning sensor to obtain multiple sample objects within the viewing angle intersection area and sample positioning data of the multiple sample objects;

[0170] Based on the sample positioning data of the multiple sample objects, an initial matching relationship between the multiple sample objects in the view angle intersection area is determined.

[0171] Viewing area information refers to scene information within the viewing area captured by the positioning sensor. This information can be presented in various forms, depending on the type and function of the positioning sensor. These forms include, but are not limited to, camera images, video frames, point clouds from infrared and radar sensors, and electrical signal spectra from radio frequency sensors.

[0172] In the embodiments of this specification, object recognition refers to the process of detecting, distinguishing, and confirming individual objects within the intersecting viewing area based on viewing area information collected by positioning sensors (such as cameras, infrared sensors, radar sensors, or radio frequency sensors). This process utilizes a variety of techniques, such as image processing, pattern recognition, and signal analysis, to extract objects with specific characteristics from complex background environments, thereby achieving effective object identification.

[0173] Optionally, obtaining the viewing area information collected by each positioning sensor in the sensor group includes the following specific steps:

[0174] According to the preset frequency, the viewing angle area information collected by each positioning sensor in the sensor group is obtained.

[0175] Please refer to the above description for details and will not go into details here.

[0176] Exemplarily, images captured by each road monitoring camera in the M-road and N-road camera groups are obtained, and based on each image, vehicle identification is performed within the viewing angle area of ​​the two road monitoring cameras. 20 sample vehicles within the viewing angle intersection area and the sample longitude and latitude and lanes of the 20 sample vehicles are obtained. Based on the sample longitude and latitude and lanes of the 20 sample vehicles, the initial matching relationship of the 20 sample vehicles within the viewing angle intersection area is determined.

[0177] In the embodiments of this specification, multiple sample objects and sample positioning data of multiple sample objects are identified through object recognition, and based on the sample positioning data of multiple sample objects, the initial matching relationship of the multiple sample objects is determined, providing support for objects, positioning data and initial matching relationships for subsequent position matching.

[0178] In an optional embodiment of the present specification, the sample object includes a sample reference object;

[0179] Correspondingly, obtaining multiple sample objects and sample positioning data of the multiple sample objects in the view intersection area includes the following specific steps:

[0180] Acquire first sample positioning data of the sample reference object within the viewing angle intersection region collected by a first positioning sensor, and second sample positioning data of the sample reference object within the viewing angle intersection region collected by a second positioning sensor, wherein the second positioning sensor is any one of the sensor group except the first positioning sensor;

[0181] Determining distance data between the first sample positioning data and the second sample positioning data based on the first sample positioning data and the second sample positioning data;

[0182] Adjusting the local coordinate system of the view intersection area collected by the second positioning sensor based on the distance data;

[0183] Based on the adjusted local coordinate system of the viewing angle intersection area collected by each second positioning sensor, sample positioning data of multiple sample objects in the viewing angle area information collected by each positioning sensor is determined.

[0184] A sample reference object is an instance of an object among multiple sample objects that serves as a reference point. By selecting a sample reference object, the local coordinate systems of different positioning sensors can be calibrated. For example, in an autonomous driving system, a road sign placed at a known location within the test area can serve as a sample reference object. This sign can be detected simultaneously by all radar sensors and onboard cameras.

[0185] The first sample positioning data of the sample reference object is data information directly collected by the first positioning sensor, which includes the position, posture, and motion state (e.g., direction, velocity, and acceleration) of the sample reference object in a specific spatial coordinate system. In the embodiments of this specification, the specific spatial coordinate system is a global coordinate system, such as longitude and latitude.

[0186] The second sample positioning data of the sample reference object is data information directly collected by the second positioning sensor, which includes the position, posture, and motion state (e.g., direction, velocity, and acceleration) of the sample reference object in a specific spatial coordinate system. In the embodiments of this specification, the specific spatial coordinate system is a global coordinate system, such as longitude and latitude.

[0187] The distance data between the first sample positioning data and the second sample positioning data is a difference value between the first sample positioning data and the second sample positioning data in a specific spatial coordinate system.

[0188] The local coordinate system of the view intersection region is an independent coordinate system established within the view intersection region and is used to represent the positioning data of objects within that region. The local coordinate system is established based on the baseline positioning data of each positioning sensor. When scene offset exists, the local coordinate system can be adjusted to ensure that the positioning data from different sensors better corresponds to the positional relationship in the real world, such as the SL coordinate system or the camera coordinate system.

[0189] Figure 7 FIG. 1 shows a schematic diagram of a local coordinate system conversion in an object positioning method provided by an embodiment of the present specification, such as Figure 7 As shown:

[0190] The first longitude and latitude [longitude, latitude] of the sample reference object within the perspective intersection area collected by the first positioning sensor, and the second longitude and latitude [longitude, latitude] of the sample reference object within the perspective intersection area collected by the second positioning sensor, are determined based on the first sample positioning data and the second sample positioning data: the horizontal and vertical distances [X, Y]; based on the distance data, the local coordinate system of the perspective intersection area collected by the second positioning sensor is adjusted: the SL coordinate system.

[0191] In the embodiments of this specification, the local coordinate system of the positioning sensor is adjusted using sample positioning data, which is applicable to all objects and is not affected by the scene, thereby improving the scalability and scene adaptability.

[0192] In an optional embodiment of the present specification, determining an initial matching relationship between multiple sample objects in a viewpoint intersection region based on sample positioning data of multiple sample objects includes the following specific steps:

[0193] Determine distance data between the plurality of sample positioning data based on the sample positioning data of the plurality of sample objects in the view intersection area collected by each positioning sensor;

[0194] Based on the distance data between the plurality of sample positioning data and a preset distance threshold, an initial matching relationship between the plurality of sample objects in the view angle intersection area is determined.

[0195] The distance data between multiple sample positioning data represents the difference between the sample positioning data of multiple sample objects. It is used to quantify the positional differences between the multiple sample objects collected by each positioning sensor. It is typically calculated using Euclidean distance, Manhattan distance, Chebyshev distance, or other suitable metrics to indicate the degree of positional deviation between two sample objects in a specific spatial coordinate system (such as latitude and longitude, three-dimensional spatial coordinates, or polar coordinates).

[0196] The preset distance threshold is a pre-set numerical limit used to determine whether the distance data between multiple sample positioning data points can be considered to be the same object entity. When the distance data between the positioning data points of a sample object obtained by different positioning sensors is less than or equal to the preset distance threshold, the sample positioning data points are determined to belong to the same object entity, and an initial matching relationship is established accordingly.

[0197] For example, in a smart transportation system, based on the location information (latitude and longitude coordinates) of the same vehicle recorded by two adjacent road monitoring cameras, the Euclidean distance between the two sets of longitude and latitude coordinates is calculated as their distance data. If the distance is less than a preset distance threshold (for example, 5 meters), it can be preliminarily determined that the two longitude and latitude coordinates are the observation results of the same sample vehicle under different camera perspectives, thereby establishing an initial matching relationship between the sample vehicles.

[0198] In the embodiments of this specification, preliminary matching of sample objects is completed by calculating distance data, and an initial matching relationship is obtained, which provides support for subsequent position matching.

[0199] See also Figure 8 , Figure 8 A flowchart of an object positioning method provided by an embodiment of this specification is shown, including the following specific steps:

[0200] Step 802: Acquire initial positioning data of the target object collected by each positioning sensor in the sensor group at multiple different times.

[0201] Step 804: Based on the position offset information, adjust each initial positioning data to obtain target positioning data of the target object at multiple different moments, wherein the position offset information is obtained based on position matching of multiple sample objects in the intersection area of ​​the viewing angles collected by each positioning sensor in the sensor group.

[0202] Step 806: Determine the target object trajectory of the target object based on the target positioning data of the target object at multiple different moments.

[0203] The embodiments of this specification are applicable to intelligent system platforms with object tracking capabilities, such as smart transportation systems, security monitoring systems, environmental monitoring systems, sports broadcasting platforms, automated warehouse management platforms, robotic navigation systems, unmanned vehicle tracking platforms, or industrial production systems based on Internet of Things technology. In these system platforms, positioning data collected by positioning sensors is used to accurately track objects and thus implement the corresponding functions.

[0204] The examples and Figure 2The embodiments of the specification are based on the same inventive concept. For the detailed contents of step 802 and step 804, please refer to the above embodiments of the specification, which will not be repeated here.

[0205] Positioning data at different moments is data information collected by positioning sensors and used to describe the position, posture, and motion state (for example, direction, speed, and acceleration) of an object at different moments in a specific spatial coordinate system. It is generally expressed in the form of time series data, including but not limited to: longitude and latitude time series, three-dimensional spatial coordinate time series, distance time series, angle time series, speed time series, acceleration time series, angular velocity time series, angular acceleration time series, direction time series, and trajectory (road) time series.

[0206] The target object trajectory of the target object is determined based on the target positioning data of the target object at multiple different moments. Specifically, the target object trajectory of the target object is obtained by fitting the motion process of the target object based on the target positioning data of the target object at multiple different moments.

[0207] For example, in a smart transportation system, multiple road monitoring cameras are pre-installed in a road scene, and four vehicles are traveling on the three-lane, one-way path of Road M. The road monitoring camera on Road M captures the viewing area and uploads the road image. Entity detection is performed on the road image using a target frame, identifying four vehicles traveling within the viewing area. Based on the baseline longitude and latitude of the road monitoring camera [longitude_0, latitude_0], the initial longitude and latitude of the four vehicles at time t0-tn and their lanes are determined. The target position offset [longitude offset, latitude offset] is added to the initial longitude and latitude of the four vehicles at time t0-tn to obtain the target longitude and latitude and their lanes at time t0-tn. Based on the target longitude and latitude and their lanes at time t0-tn, the motion processes of the four vehicles are fitted to obtain the vehicle motion trajectories of the four vehicles at time t0-tn.

[0208] In the embodiments of this specification, based on position matching of multiple sample objects in the intersecting area of ​​the viewing angles collected by each positioning sensor, the overall positioning offset information of the positioning sensor is estimated, and the initial positioning data collected under the scene offset condition is eliminated, thereby improving the accuracy of the object positioning result and the object positioning effect. The target object trajectory is then determined through the highly accurate target positioning data, thereby improving the accuracy of the object tracking result and the object tracking effect.

[0209] In an optional embodiment of this specification, step 802 includes the following specific steps:

[0210] Acquire viewing area information collected by each positioning sensor in the sensor group at multiple different moments, wherein the viewing area information is scene information within the viewing area collected by each positioning sensor;

[0211] Based on the viewing angle area information, object recognition is performed within the viewing angle area of ​​each positioning sensor to obtain initial positioning data of the target object at multiple different times.

[0212] The examples and Figure 2 The embodiments of the specification are based on the same inventive concept. For details, please refer to the above embodiments of the specification, which will not be repeated here.

[0213] In the embodiments of this specification, multiple target objects and initial positioning data of the multiple target objects are identified and obtained through object recognition, which provides object and positioning data support for subsequent object positioning.

[0214] In an optional embodiment of this specification, before step 806, the following specific steps are further included:

[0215] Get the object identification information of the target object;

[0216] Correspondingly, step 806 includes the following specific steps:

[0217] Based on the object identification information of the target object and the target positioning data at multiple different moments, the motion process of the target object is fitted to obtain the target object trajectory of the target object.

[0218] The object identification information of the target object is identification information used to uniquely or clearly identify and distinguish different target objects, including but not limited to: target frame position, target frame size, object identification (object ID). For example, in the intelligent transportation system, suppose there is a vehicle with a license plate number of ". For example, in the intelligent transportation system, the vehicle is traveling in the field of view of a road monitoring camera. The object identification information of the target object may include: target frame position. In a certain frame image, the position of the vehicle is described as (x=100, y=200) to (x=300, y=400), which represents the coordinates of the upper left corner and lower right corner of the bounding box of the vehicle in the frame image. Target frame size: 200 pixels wide and 200 pixels high, indicating the size of the vehicle in the image. Object identification (object ID): ":" indicates the size of the vehicle in the image. The system assigns this vehicle a unique internal identifier. No matter which positioning sensor's field of view it appears in, as long as it is identified as the same vehicle, it will be recorded with the same object ID.

[0219] Based on the object identification information of the target object and the target positioning data at multiple different times, the motion process of the target object is fitted to obtain the target object trajectory of the target object. The fitting methods include but are not limited to: Kalman filtering, polynomial fitting and hidden Markov model.

[0220] For example, the object ID of "vehicle_001" is used to associate the target longitude and latitude at time t0-tn, and the target longitude and latitude at time t0-tn are input into the Kalman filter. The filter combines the target longitude and latitude at the current moment with the target longitude and latitude at the previous moment, and outputs the target estimate of the vehicle position and speed. The above process is repeated to continuously apply the Kalman filter, and gradually fit the continuous and smooth vehicle motion trajectories of the four vehicles on Road M from time t0-tn.

[0221] In the embodiments of this specification, based on the object identification information of the target object and the target positioning data at multiple different times, the movement process of the target object is fitted to obtain the target object trajectory of the target object, thereby ensuring the continuity and consistency of object tracking, further improving the accuracy of object tracking, and further improving the object tracking effect.

[0222] Regarding the above Figure 8 Instruction Manual Examples, Figure 9 FIG. 1 shows a flow chart of an object tracking method provided by an embodiment of the present specification. Figure 9 As shown:

[0223] First, object detection: obtain the viewing area information collected by each positioning sensor in the sensor group at multiple different times; based on the viewing area information, perform object recognition in the viewing area of ​​each positioning sensor to determine the target object in the viewing area.

[0224] Next, object identification: obtain the object identification information of the target object.

[0225] Then, object positioning: obtain the initial positioning data of the target object at multiple different times.

[0226] Then, positioning adjustment: based on the position offset information, the initial positioning data at multiple different moments are adjusted to obtain the target positioning data of the target object at multiple different moments.

[0227] Finally, object tracking, based on the target positioning data at multiple different times, as well as the target box position, target box size, target longitude and latitude and object identification (object ID), fits the motion process of the target object to obtain the target object trajectory.

[0228] See also Figure 10 , Figure 10A flowchart of a vehicle tracking method in a road network scenario provided by one embodiment of this specification is shown. The method is applied to a cloud-side device and includes the following specific steps:

[0229] Step 1002: Acquire initial positioning data of the target vehicle at multiple different times collected by each camera in the camera group of the target road scene.

[0230] Step 1004: Based on the position offset information, adjust each initial positioning data to obtain the target positioning data of the target vehicle at multiple different times, wherein the position offset information is obtained by position matching multiple sample vehicles within the perspective intersection area captured by each camera in the camera group, and the perspective intersection area is the area where the perspective areas of each camera in the camera group intersect.

[0231] Step 1006: Determine a target vehicle trajectory of the target vehicle in the target road scene based on the target positioning data of the target vehicle at multiple different moments.

[0232] The embodiments of this specification are applied to an intelligent transportation system with a vehicle tracking function in a road network scenario, and the system is deployed on a cloud-side device.

[0233] The examples and Figure 2 Instruction Manual Examples Figure 8 The embodiments of the specification are based on the same inventive concept. For the detailed contents of step 1002 and step 1006, please refer to the above embodiments of the specification, which will not be repeated here.

[0234] In the embodiments of this specification, based on position matching of multiple sample vehicles in the intersection area of ​​the perspectives captured by each camera, the overall positioning offset information of the camera is estimated, and the initial positioning data collected under the scene offset condition is eliminated, thereby improving the accuracy of the vehicle positioning result and the vehicle positioning effect. The target vehicle trajectory is then determined through high-accuracy target positioning data, thereby improving the accuracy of the vehicle tracking result and improving the vehicle tracking effect in the road network scene.

[0235] The following combined Figure 11 , taking the application of the object tracking method provided in this specification in the global road network scenario as an example, the object tracking method is further explained. Figure 11 A flowchart of a vehicle tracking method applied to a global road network scenario provided by an embodiment of this specification is shown, including the following specific steps:

[0236] Step 1102: Obtain road images captured by each camera in the global camera group at a preset frequency.

[0237] A global camera group, consisting of three cameras, A and B, is pre-configured to monitor city roads from different angles. The default frequency is to capture road images every five minutes. At the preset frequency (for example, every five minutes), each camera, A or B, will capture images of the road within its field of view.

[0238] Step 1104: Based on the road image, perform vehicle recognition within the viewing angle area of ​​each camera, and obtain multiple sample vehicles within the viewing angle intersection area and sample positioning data of the multiple sample vehicles.

[0239] In the intersection area of ​​camera views A and B, vehicles 1, 2, and 3 are identified, and the coordinate positions of these vehicles in their respective images are recorded.

[0240] Step 1106: Based on the sample positioning data of the multiple sample vehicles, determine the initial matching relationship of the multiple sample vehicles in the perspective intersection area, wherein the multiple sample vehicles include multiple first sample vehicles in the perspective intersection area captured by the first camera and multiple second sample vehicles in the perspective intersection area captured by the second camera, and there is an initial matching relationship between any first sample vehicle and at least one second sample vehicle, the first camera is any one in the global camera group, and the second camera is any one in the global camera group except the first camera.

[0241] Vehicle 1 appears in both the images of camera A and camera B, so it is determined that there is an initial matching relationship between the two cameras.

[0242] Step 1108: Randomly select a first sample vehicle from multiple first sample vehicles, randomly select a second sample vehicle from at least one second sample vehicle that has an initial matching relationship with the first sample vehicle, and determine the initial matching relationship between the first sample vehicle and the second sample vehicle as the target matching relationship, delete the first sample vehicle from the multiple first sample vehicles, and return to execute the step of randomly selecting a first sample vehicle from the multiple first sample vehicles until the target matching relationship between each first sample vehicle and the second sample vehicle is obtained.

[0243] Step 1110 : When the number of target matching relationships of the plurality of sample vehicles reaches a preset threshold, determine the target matching relationships of the plurality of sample vehicles as sample matching relationships of the plurality of sample vehicles.

[0244] Step 1112: If the number of target matching relationships of the plurality of sample vehicles does not reach a preset threshold, return to step 1108 for execution.

[0245] First, randomly select vehicle 1 as the first sample vehicle. Combine this with vehicle 1 in camera B's view as the second sample vehicle. Set the relationship between the two as the target matching relationship. Then remove vehicle 1 from the first sample vehicle list. Repeat this process until all vehicles have a target matching relationship. If the number of target matching relationships does not cover all vehicles in the intersection area, continue sampling until a target matching relationship is obtained between each first sample vehicle and the second sample vehicle.

[0246] Step 1114: Based on the sample matching relationship of the multiple sample vehicles, calculate the offset of the sample positioning data of the multiple sample vehicles, and determine the position offset based on the offset of each sample positioning data.

[0247] Calculate the coordinate difference of each vehicle under the A and B cameras.

[0248] Step 1116: Acquire road images captured by each camera in the global camera group at multiple different times.

[0249] Step 1118: Based on the road image, perform vehicle recognition within the viewing area of ​​each camera to obtain initial positioning data of the target vehicle at multiple different times.

[0250] Cameras A and B collect several consecutive road images within 5 minutes and identify the initial coordinate position of the tracking vehicle 4 at each moment.

[0251] Step 1120: Based on the position offset, adjust the initial positioning data to obtain target positioning data of the target vehicle.

[0252] The initial coordinate position is added to the coordinate difference to obtain the target coordinate position of vehicle 3.

[0253] Step 1122: Based on the vehicle identification information of the target vehicle and the target positioning data at multiple different times, the motion process of the target object is fitted to obtain the global vehicle trajectory of the target vehicle.

[0254] Based on the identification information of vehicle 4 and the adjusted target coordinate positions at multiple different times, the trajectory fitting algorithm is used to calculate the global vehicle trajectory of vehicle 4 in the entire monitoring area, thereby achieving continuous tracking of the target vehicle.

[0255] The embodiments of this specification and the embodiments of the above specification are based on the same inventive concept. For the specific content of each step in the embodiments of this specification, please refer to the embodiments of the above specification and will not be repeated here.

[0256] In the embodiments of this specification, in order to address the problem that scene offset causes target positioning offset, resulting in poor cross-camera vehicle global tracking, the position offset is estimated by adding a Hungarian matching algorithm. Scene offset usually causes an overall offset in the positioning data. Although the distance between the same target in the two cameras becomes larger, the relative position of the vehicle remains unchanged. This feature is suitable for target matching using the Hungarian matching algorithm. The average distance between successfully matched targets can be approximated as the offset of the target. The estimated offset is used in the Hungarian matching optimization module to correct the positioning data of the public area, and the distance data is used to associate the vehicles. The Hungarian matching algorithm is used to monitor and alarm for scene offset problems.

[0257] Figure 12 FIG. 1 shows a front-end schematic diagram of a vehicle tracking method applied to a global road network scenario provided by an embodiment of this specification, such as Figure 12 As shown:

[0258] The main function of the front-end interface is to generate and display vehicle trajectories. Specifically, the interface consists of the following parts: a title area at the top, which displays the "front-end interface" and a return control. A graphical area in the middle displays a trajectory diagram composed of dashed lines and arrows, indicating the movement of the vehicle between different locations. A data table area at the bottom lists the current camera position, the previous camera position, the next camera position, and the corresponding timestamps.

[0259] Users can query vehicle trajectory information by entering a license plate number and clicking the "Search" control. The system then generates a corresponding vehicle trajectory based on the entered license plate number and displays it in a data table. Each location is associated with a timestamp, allowing users to understand the vehicle's position changes at different points in time. In addition, the interface provides a "Generate Vehicle Track" control that users can click to generate a new vehicle track. The generated track is displayed in the graphics area, and the corresponding information is listed in the data table.

[0260] Corresponding to the above method embodiment, this specification also provides an object tracking system embodiment, Figure 13 FIG. 1 shows a schematic diagram of the structure of an object tracking system provided by an embodiment of this specification. Figure 13 As shown, the system includes:

[0261] The object detection unit 1302 is configured to obtain viewing area information collected by each positioning sensor in the sensor group at multiple different times; based on the viewing area information, perform object recognition within the viewing area of ​​each positioning sensor to determine the target object within the viewing area;

[0262] The object identification unit 1304 is used to obtain the object identification information of the target object;

[0263] The object positioning unit 1306 is configured to obtain initial positioning data of the target object at multiple different time instants, and adjust each initial positioning data based on position offset information to obtain target positioning data of the target object at multiple different time instants, wherein the position offset information is obtained by position matching multiple sample objects within an intersecting view area captured by each positioning sensor, where the view area intersects the view areas of each positioning sensor in the sensor group;

[0264] The object tracking unit 1308 is configured to fit the motion process of the target object based on the object identification information of the target object and the target positioning data at multiple different moments, so as to obtain the target object trajectory of the target object.

[0265] In an optional embodiment of the present specification, the system further includes: an object matching unit, configured to obtain sample positioning data of a plurality of sample objects within an intersecting area of ​​viewing angles collected by each positioning sensor in the sensor group, and an initial matching relationship between the plurality of sample objects within the intersecting area of ​​viewing angles; based on the initial matching relationship, position matching is performed on the plurality of sample objects to obtain a sample matching relationship between the plurality of sample objects; and position offset information is determined based on the sample positioning data and the sample matching relationship of the plurality of sample objects.

[0266] In the embodiments of the present specification, based on position matching of multiple sample objects within the intersecting area of ​​the viewing angles collected by each positioning sensor, the overall positioning offset information of the positioning sensor is estimated. After the viewing area information at multiple different moments is collected by each positioning sensor, object recognition is performed within the viewing area of ​​each positioning sensor based on the viewing area information to determine the target object within the viewing area. The initial positioning data collected under the scene offset condition is eliminated by the position offset information, thereby improving the accuracy of the object positioning result and the object positioning effect. Furthermore, the trajectory of the target object is determined by the highly accurate target positioning data and object identification information, thereby improving the accuracy of the object tracking result and the object tracking effect.

[0267] The above is a schematic diagram of an object tracking system according to this embodiment. It should be noted that the technical solution of this object tracking system and the technical solution of the object tracking method described above are based on the same concept. For details not described in detail in the technical solution of the object tracking system, please refer to the description of the technical solution of the object tracking method described above.

[0268] Corresponding to the above method embodiment, this specification also provides an object positioning device embodiment, Figure 14 FIG. 1 shows a schematic diagram of the structure of an object positioning device provided by an embodiment of this specification. Figure 14 As shown, the device includes:

[0269] A first acquisition module 1402 is configured to acquire initial positioning data of a target object collected by a first positioning sensor;

[0270] The first adjustment module 1404 is configured to adjust the initial positioning data based on the position offset information to obtain target positioning data of the target object, wherein the position offset information is obtained by position matching a plurality of sample objects within an intersection area of ​​viewing angles collected by each positioning sensor in the sensor group, the sensor group including the first positioning sensor, and the intersection area of ​​viewing angles being an area where viewing angles of each positioning sensor in the sensor group intersect.

[0271] Optionally, the device also includes: an offset information determination module, configured to obtain sample positioning data of multiple sample objects within the perspective intersection area collected by each positioning sensor in the sensor group, and an initial matching relationship between the multiple sample objects within the perspective intersection area; based on the initial matching relationship, perform position matching on the multiple sample objects to obtain sample matching relationships of the multiple sample objects; and determine position offset information based on the sample positioning data and sample matching relationships of the multiple sample objects.

[0272] Optionally, the offset information determination module is further configured to: perform position matching on multiple sample objects based on the initial matching relationship to obtain a target matching relationship of the multiple sample objects; when the number of target matching relationships of the multiple sample objects reaches a preset threshold, determine the target matching relationship of the multiple sample objects as a sample matching relationship of the multiple sample objects.

[0273] Optionally, the multiple sample objects include multiple first sample objects within an intersecting area of ​​viewing angles captured by the first positioning sensor and multiple second sample objects within an intersecting area of ​​viewing angles captured by the second positioning sensor, an initial matching relationship exists between any first sample object and at least one second sample object, and the second positioning sensor is any one of the sensor group except the first positioning sensor;

[0274] Correspondingly, the offset information determination module is further configured to: arbitrarily extract a first sample extraction object from multiple first sample objects; arbitrarily extract a second sample extraction object from at least one second sample object that has an initial matching relationship with the first sample extraction object, and determine the initial matching relationship between the first sample extraction object and the second sample extraction object as a target matching relationship; delete the first sample extraction object from the multiple first sample objects, and return to execute the step of arbitrarily extracting a first sample extraction object from the multiple first sample objects, until a target matching relationship is obtained between each first sample object and the second sample object.

[0275] Optionally, the offset information determining module is further configured to: calculate offset information of sample positioning data of multiple sample objects based on sample matching relationships of the multiple sample objects; and determine position offset information based on the offset information of each sample positioning data.

[0276] Optionally, the offset information determination module is further configured to: obtain viewing area information collected by each positioning sensor in the sensor group, wherein the viewing area information is scene information within the viewing area collected by each positioning sensor; based on the viewing area information, perform object recognition within the viewing area of ​​each positioning sensor, obtain multiple sample objects within the viewing intersection area and sample positioning data of the multiple sample objects; and determine the initial matching relationship of the multiple sample objects within the viewing intersection area based on the sample positioning data of the multiple sample objects.

[0277] Optionally, the sample object includes a sample reference object; correspondingly, the offset information determination module is further configured to: obtain first sample positioning data of the sample reference object within the perspective intersection area collected by the first positioning sensor, and second sample positioning data of the sample reference object within the perspective intersection area collected by the second positioning sensor, wherein the second positioning sensor is any one of the sensor group except the first positioning sensor; based on the first sample positioning data and the second sample positioning data, determine the distance data between the first sample positioning data and the second sample positioning data; based on the distance data, adjust the local coordinate system of the perspective intersection area collected by the second positioning sensor; based on the adjusted local coordinate system of the perspective intersection area collected by each second positioning sensor, determine the sample positioning data of multiple sample objects within the perspective area information collected by each positioning sensor.

[0278] Optionally, the offset information determination module is further configured to: determine distance data between the multiple sample positioning data based on the sample positioning data of the multiple sample objects in the perspective intersection area collected by each positioning sensor; and determine the initial matching relationship of the multiple sample objects in the perspective intersection area based on the distance data between the multiple sample positioning data and a preset distance threshold.

[0279] In the embodiments of this specification, based on position matching of multiple sample objects in the intersection area of ​​the perspectives collected by each positioning sensor, the overall positioning offset information of the positioning sensor is estimated, and this is used to eliminate the initial positioning data collected under the scene offset condition, thereby improving the accuracy of the object positioning result and improving the object positioning effect.

[0280] The above is a schematic diagram of an object positioning device according to this embodiment. It should be noted that the technical solution of the object positioning device and the technical solution of the object positioning method described above are based on the same concept. For details not described in detail in the technical solution of the object positioning device, please refer to the description of the technical solution of the object positioning method described above.

[0281] Corresponding to the above method embodiment, this specification also provides an object tracking device embodiment, Figure 15 FIG. 1 shows a schematic diagram of the structure of an object tracking device provided by an embodiment of this specification. Figure 15 As shown, the device includes:

[0282] The second acquisition module 1502 is configured to acquire initial positioning data of the target object collected by each positioning sensor in the sensor group at multiple different times;

[0283] A second adjustment module 1504 is configured to adjust each initial positioning data based on the position offset information to obtain target positioning data of the target object at multiple different time instants, wherein the position offset information is obtained by position matching a plurality of sample objects within an intersecting area of ​​viewing angles collected by each positioning sensor in the sensor group, where the intersecting area of ​​viewing angles is an area where the viewing angles of each positioning sensor in the sensor group intersect;

[0284] The second tracking module 1506 is configured to determine a target object trajectory of the target object based on target positioning data of the target object at multiple different moments.

[0285] Optionally, the second acquisition module 1502 is further configured to: obtain viewing area information collected by each positioning sensor in the sensor group at multiple different moments, wherein the viewing area information is scene information within the viewing area collected by each positioning sensor; based on the viewing area information, perform object recognition within the viewing area of ​​each positioning sensor to obtain initial positioning data of the target object at multiple different moments.

[0286] Optionally, the apparatus further comprises: an object tracking module configured to obtain object identification information of the target object;

[0287] Correspondingly, the second tracking module 1506 is further configured to fit the motion process of the target object based on the object identification information of the target object and the target positioning data at multiple different moments to obtain the target object trajectory of the target object.

[0288] In the embodiments of this specification, based on position matching of multiple sample objects in the intersecting area of ​​the viewing angles collected by each positioning sensor, the overall positioning offset information of the positioning sensor is estimated, and the initial positioning data collected under the scene offset condition is eliminated, thereby improving the accuracy of the object positioning result and the object positioning effect. The target object trajectory is then determined through the highly accurate target positioning data, thereby improving the accuracy of the object tracking result and the object tracking effect.

[0289] The above is a schematic diagram of an object tracking device according to this embodiment. It should be noted that the technical solution of this object tracking device and the technical solution of the object tracking method described above are based on the same concept. For details not described in detail in the technical solution of the object tracking device, please refer to the description of the technical solution of the object tracking method described above.

[0290] Corresponding to the above method embodiment, this specification also provides an embodiment of a vehicle tracking device in a road network scenario. Figure 16 FIG1 shows a schematic diagram of the structure of a vehicle tracking device in a road network scenario provided by an embodiment of this specification. Figure 16 As shown, the device is applied to cloud-side equipment and includes:

[0291] The third acquisition module 1602 is configured to obtain initial positioning data of the target vehicle at multiple different times collected by each camera in the camera group of the target road scene;

[0292] A third adjustment module 1604 is configured to adjust each initial positioning data based on the position offset information to obtain target positioning data of the target vehicle at multiple different time instants, wherein the position offset information is obtained by position matching multiple sample vehicles within an intersecting area of ​​view angles captured by each camera in the camera group, where the intersecting area of ​​view angles is an area where the view angles of each camera in the camera group intersect;

[0293] The third tracking module 1606 is configured to determine a target vehicle trajectory of the target vehicle in the target road scene based on the target positioning data of the target vehicle at multiple different moments.

[0294] In the embodiments of this specification, based on position matching of multiple sample vehicles in the intersection area of ​​the perspectives captured by each camera, the overall positioning offset information of the camera is estimated, and the initial positioning data collected under the scene offset condition is eliminated, thereby improving the accuracy of the vehicle positioning result and the vehicle positioning effect. The target vehicle trajectory is then determined through high-accuracy target positioning data, thereby improving the accuracy of the vehicle tracking result and improving the vehicle tracking effect in the road network scene.

[0295] The above is a schematic diagram of a vehicle tracking device in a road network scenario according to this embodiment. It should be noted that the technical solution of this vehicle tracking device in a road network scenario is based on the same concept as the technical solution of the vehicle tracking method in a road network scenario described above. For details not described in detail in the technical solution of the vehicle tracking device in a road network scenario, please refer to the description of the technical solution of the vehicle tracking method in a road network scenario described above.

[0296] Figure 1717 shows a block diagram of a computing device according to an embodiment of the present disclosure. Components of the computing device 1700 include, but are not limited to, a memory 1710 and a processor 1720. The processor 1720 is connected to the memory 1710 via a bus 1730, and a database 1750 is used to store data.

[0297] The computing device 1700 also includes an access device 1740 that enables the computing device 1700 to communicate via one or more networks 1760. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 1740 may include one or more of any type of network interface (e.g., a network interface card (NIC)) of wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a near field communication (NFC).

[0298] In one embodiment of the present specification, the above components of the computing device 1700 and Figure 17 Other components not shown in the figure may also be connected to each other, for example, via a bus. Figure 17 The computing device structure block diagram shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art may add or replace other components as needed.

[0299] Computing device 1700 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, personal digital assistant, laptop computer, notebook computer, netbook computer, etc.), a mobile phone (e.g., a smartphone), a wearable computing device (e.g., a smartwatch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or personal computer (PC). Computing device 1700 may also be a mobile or stationary server.

[0300] Among them, the processor 1720 is used to execute the following computer program / instruction, which, when executed by the processor, implements the steps of the above-mentioned object positioning method, object tracking method or vehicle tracking method in a road network scenario.

[0301] The above is a schematic diagram of a computing device according to this embodiment. It should be noted that the technical solution of this computing device is based on the same concept as the technical solutions of the aforementioned object positioning method, object tracking method, and vehicle tracking method in a road network scenario. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solutions of the aforementioned object positioning method, object tracking method, or vehicle tracking method in a road network scenario.

[0302] An embodiment of the present specification further provides a computer-readable storage medium storing a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned object positioning method, object tracking method, or vehicle tracking method in a road network scenario.

[0303] The above is an illustrative embodiment of a computer-readable storage medium. It should be noted that the technical solution of this storage medium is based on the same concept as the technical solutions of the aforementioned object positioning method, object tracking method, and vehicle tracking method in a road network scenario. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solutions of the aforementioned object positioning method, object tracking method, or vehicle tracking method in a road network scenario.

[0304] An embodiment of the present specification further provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned object positioning method, object tracking method, or vehicle tracking method in a road network scenario.

[0305] The above is an illustrative solution of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product is based on the same concept as the technical solutions of the aforementioned object positioning method, object tracking method, and vehicle tracking method in a road network scenario. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solutions of the aforementioned object positioning method, object tracking method, or vehicle tracking method in a road network scenario.

[0306] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0307] The computer instructions include computer program code, which may be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0308] It should be noted that for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0309] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0310] The preferred embodiments disclosed above are intended only to help illustrate this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made based on the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that those skilled in the art can better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A method for object positioning, comprising: Acquire initial positioning data of the target object collected by the first positioning sensor; Based on the position offset information, the initial positioning data is adjusted to obtain target positioning data of the target object, wherein the position offset information is obtained based on position matching of multiple sample objects within an intersection area of ​​viewing angles collected by each positioning sensor in the sensor group, the sensor group includes the first positioning sensor, and the intersection area of ​​viewing angles is an area where viewing angles of each positioning sensor in the sensor group intersect.

2. The method according to claim 1, before adjusting the initial positioning data based on the position offset information to obtain the target positioning data of the target object, further comprising: Acquire sample positioning data of a plurality of sample objects in a viewing angle intersection region collected by each positioning sensor in the sensor group, as well as initial matching relationships of the plurality of sample objects in the viewing angle intersection region; Based on the initial matching relationship, position matching is performed on the multiple sample objects to obtain a sample matching relationship between the multiple sample objects; The position offset information is determined based on the sample positioning data of the multiple sample objects and the sample matching relationship.

3. The method according to claim 2, wherein the performing position matching on the plurality of sample objects based on the initial matching relationship to obtain the sample matching relationship of the plurality of sample objects comprises: Based on the initial matching relationship, position matching is performed on the multiple sample objects to obtain a target matching relationship between the multiple sample objects; When the number of the target matching relationships of the multiple sample objects reaches a preset threshold, the target matching relationships of the multiple sample objects are determined to be the sample matching relationships of the multiple sample objects.

4. The method according to claim 3, wherein the plurality of sample objects include a plurality of first sample objects within the intersecting area of ​​the viewing angles captured by the first positioning sensor and a plurality of second sample objects within the intersecting area of ​​the viewing angles captured by the second positioning sensor, an initial matching relationship exists between any of the first sample objects and at least one of the second sample objects, and the second positioning sensor is any one of the sensor group except the first positioning sensor; The performing position matching on the plurality of sample objects based on the initial matching relationship to obtain a target matching relationship of the plurality of sample objects includes: Randomly extract a first sample extraction object from the multiple first sample objects; Randomly extracting a second sample object from at least one second sample object that has the initial matching relationship with the first sample object, and determining the initial matching relationship between the first sample object and the second sample object as the target matching relationship; The first sample extraction object is deleted from the multiple first sample objects, and the step of randomly extracting a first sample extraction object from the multiple first sample objects is returned to be executed until the target matching relationship between each of the first sample objects and the second sample objects is obtained.

5. The method according to claim 2, wherein determining the position offset information based on the sample positioning data and the sample matching relationship of the plurality of sample objects comprises: Calculating offset information of sample positioning data of the multiple sample objects based on the sample matching relationships of the multiple sample objects; The position offset information is determined based on the offset information of the sample positioning data of each of the sample objects.

6. The method according to claim 2, wherein obtaining sample positioning data of a plurality of sample objects within a viewing angle intersection region collected by each positioning sensor in the sensor group, and initial matching relationships of the plurality of sample objects within the viewing angle intersection region, comprises: Acquire viewing area information collected by each positioning sensor in the sensor group, wherein the viewing area information is scene information within the viewing area collected by each positioning sensor; Based on the viewing angle area information, performing object recognition within the viewing angle area of ​​each positioning sensor, and acquiring the plurality of sample objects within the viewing angle intersection area and the sample positioning data of the plurality of sample objects; Based on the sample positioning data of the multiple sample objects, the initial matching relationship between the multiple sample objects in the perspective intersection area is determined.

7. The method of claim 6, wherein the sample object comprises a sample reference object; The acquiring of a plurality of sample objects within the view intersection area and sample positioning data of the plurality of sample objects includes: Acquire first sample positioning data of the sample reference object within the viewing angle intersection area collected by the first positioning sensor, and second sample positioning data of the sample reference object within the viewing angle intersection area collected by the second positioning sensor, wherein the second positioning sensor is any one of the sensor group except the first positioning sensor; Determining distance data between the first sample positioning data and the second sample positioning data based on the first sample positioning data and the second sample positioning data; Adjusting the local coordinate system of the view intersection area collected by the second positioning sensor based on the distance data; Based on the adjusted local coordinate system of the viewing angle intersection area collected by each of the second positioning sensors, the sample positioning data of multiple sample objects in the viewing angle area information collected by each of the positioning sensors are determined.

8. The method according to claim 6 or 7, wherein determining the initial matching relationship between the plurality of sample objects in the view intersection area based on the sample positioning data of the plurality of sample objects comprises: Determining distance data between the plurality of sample positioning data based on the sample positioning data of the plurality of sample objects in the view intersection area collected by the positioning sensors; The initial matching relationship between the plurality of sample objects in the view angle intersection area is determined based on the distance data between the plurality of sample positioning data and a preset distance threshold.

9. A method for object tracking, comprising: Obtaining initial positioning data of the target object collected by each positioning sensor in the sensor group at multiple different times; Adjusting each of the initial positioning data based on the position offset information to obtain target positioning data of the target object at multiple different moments, wherein the position offset information is obtained by position matching multiple sample objects within an intersection area of ​​viewing angles collected by each positioning sensor in the sensor group, the intersection area of ​​viewing angles being an area where viewing angles of each positioning sensor in the sensor group intersect; A target object trajectory of the target object is determined based on target positioning data of the target object at multiple different moments.

10. The method according to claim 9, wherein obtaining initial positioning data of the target object collected by each positioning sensor in the sensor group at multiple different times comprises: Acquire the viewing area information at the plurality of different moments collected by each positioning sensor in the sensor group, wherein the viewing area information is scene information within the viewing area collected by each positioning sensor; Based on the viewing angle area information, object recognition is performed within the viewing angle area of ​​each positioning sensor to obtain initial positioning data of the target object at the multiple different time moments.

11. The method according to claim 9 or 10, before determining the target object trajectory of the target object based on the target positioning data of the target object at multiple different moments, comprising: Obtaining object identification information of the target object; The determining the target object trajectory of the target object based on the target positioning data of the target object at multiple different moments includes: Based on the object identification information of the target object and the target positioning data at the multiple different moments, the motion process of the target object is fitted to obtain the target object trajectory of the target object.

12. A vehicle tracking method in a road network scenario, applied to a cloud-side device, comprising: Obtaining initial positioning data of the target vehicle at multiple different times collected by each camera in a camera group of the target road scene; Adjusting each initial positioning data based on the position offset information to obtain target positioning data of the target vehicle at the multiple different time instants, wherein the position offset information is obtained by position matching multiple sample vehicles within an intersection area of ​​viewing angles captured by each camera in the camera group, where the viewing angle intersection area is an area where viewing angles of each camera in the camera group intersect; A target vehicle trajectory of the target vehicle in the target road scene is determined based on the target positioning data of the target vehicle at the multiple different time instants.

13. An object tracking system comprising: An object detection unit, configured to obtain visual area information collected by each positioning sensor in the sensor group at multiple different moments; Based on the viewing angle area information, performing object recognition within the viewing angle area of ​​each positioning sensor to determine the target object within the viewing angle area; An object identification unit, configured to obtain object identification information of the target object; an object positioning unit, configured to obtain initial positioning data of the target object at the plurality of different moments, and adjust each initial positioning data based on position offset information to obtain target positioning data of the target object at the plurality of different moments, wherein the position offset information is obtained by position matching a plurality of sample objects within an intersection region of viewing angles collected by each of the positioning sensors, the intersection region being an area where the viewing angles of each of the positioning sensors in the sensor group intersect; The object tracking unit is used to fit the motion process of the target object based on the object identification information of the target object and the target positioning data at multiple different moments to obtain the target object trajectory of the target object.

14. The system according to claim 13, further comprising: an object matching unit, configured to obtain sample positioning data of a plurality of sample objects within an intersecting area of ​​viewing angles collected by each positioning sensor in the sensor group, and an initial matching relationship between the plurality of sample objects within the intersecting area of ​​viewing angles; Based on the initial matching relationship, position matching is performed on the multiple sample objects to obtain a sample matching relationship between the multiple sample objects; The position offset information is determined based on the sample positioning data of the multiple sample objects and the sample matching relationship.

15. A computing device comprising: memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer program / instructions are executed by the processor, the steps of the method according to any one of claims 1 to 12 are implemented.

16. A computer-readable storage medium storing a computer program / instruction, wherein the computer program / instruction, when executed by a processor, implements the steps of the method according to any one of claims 1 to 12.

17. A computer program product comprising a computer program / instruction, which implements the steps of the method according to any one of claims 1 to 12 when executed by a processor.

Images