Spatial perception observation information quality evaluation method and system, terminal and medium

Abstract

The invention relates to the field of robot navigation, and particularly discloses a spatial perception observation information quality evaluation method and system, a terminal and a medium, and the method comprises the steps: converting the measurement information of each dimension into a normalization probability according to the measurement target of each dimension; according to the concept of entropy in the information theory, the normalized probability of each dimension is converted into a computable entropy value; a downstream task weight coefficient is introduced for the entropy value of each dimension, the multi-dimensional entropy value after the downstream task weight coefficient is introduced is mapped into a physical interpretable confidence coefficient in a [0, 1] interval through an exponential attenuation function, and the confidence coefficient explicitly represents the uncertainty of observation information of each dimension. According to the method, a space perception quality evaluation system with mathematics interpretability, scene adaptation and real-time calculation is formed, multi-source uncertainty is mapped into executable confidence, and multi-dimensional accurate evaluation of intelligent agent space perception observation information is achieved.

Description

Technical Field

[0001] This invention relates to the field of robot navigation, specifically to a method, system, terminal, and medium for assessing the quality of spatial perception and observation information. Background Technology

[0002] In the navigation and operating system of mobile robots, the accuracy of environmental perception information directly determines the success or failure of task execution. The perception system bears the core function of interpreting spatial topological relationships, object geometric attributes, and dynamic scene evolution, and the reliability of its output is crucial in every stage of localization and mapping, path planning, and execution control. Once there is an unquantified deviation in the perception information, it will trigger a cascading failure of the decision chain: a small error in spatial localization may lead to collisions in path planning, deviations in the calculation of the target object's pose will cause the robotic arm to fail to grasp, and misjudgment of the trajectory of dynamic obstacles can even lead to catastrophic safety risks.

[0003] Current robotic systems rely on perceptual quality assessment mechanisms with fundamental flaws. Traditional methods often employ rigid judgment rules, such as filtering point cloud data using fixed distance thresholds. This type of binary judgment cannot differentiate the fault tolerance requirements of different task scenarios. Centimeter-level errors that are tolerable in navigation tasks can lead to fatal failures in precision assembly scenarios. More seriously, mainstream methods mix and process multiple sources of uncertainty, including sensor noise (such as multipath interference from depth cameras), environmental disturbances (such as feature loss due to sudden changes in illumination), and algorithmic limitations (such as blurred segmentation edges), ultimately outputting only simple Boolean values ​​or general probability scores. This approach masks the physical root causes of errors, making it difficult for engineers to trace the source of failure—whether it stems from sensor hardware defects, environmental interference, or algorithmic logic flaws—when the system fails.

[0004] Deep learning-driven perceptual models further exacerbate the interpretability crisis. Although end-to-end architectures can output probabilistic confidence scores, their decision-making process is like a black box: the same object may exhibit drastically fluctuating confidence scores under different poses, and a sudden low confidence score in a certain area cannot be correlated with perceptual degradation in a specific dimension (such as a surge in depth noise or geometric structural breakage). This lack of interpretability leads to a deadlock in fault diagnosis, forcing system maintenance to rely on trial and error rather than scientific attribution.

[0005] A deeper contradiction lies in the disconnect between evaluation strategies and task objectives. Mobile robots have fundamentally different reliability requirements in different scenarios: navigation tasks focus on the spatial connectivity confidence of channel width estimation, precision assembly requires sub-millimeter-level geometric consistency assurance, and dynamic grasping relies on the temporal stability evaluation of motion trajectories. However, existing technologies use static, uniform parameter configurations to address all scenarios. This "one-size-fits-all" evaluation model leads to rigid system performance in complex environments.

[0006] The aforementioned deficiencies all point to a core contradiction: the perception system lacks a reliability assessment mechanism that is continuously quantifiable, physically interpretable, and task-adaptive. This deficiency forces the decision-making layer to make risky, either-or actions under incomplete information, resulting in a persistently high failure rate for the agent in complex scenarios. When the coupling effect of environmental disturbances and algorithmic limitations cannot be explicitly represented, the system can neither proactively avoid high-risk operations nor achieve accurate attribution and optimization after failures, ultimately forming a vicious cycle of "perception misjudgment, operational failure, human intervention, and defect cover-up." Summary of the Invention

[0007] To address the aforementioned issues, this invention provides a method, system, terminal, and medium for assessing the quality of spatial perception observation information. This forms a mathematically interpretable, scenario-adaptable, and real-time computable spatial perception quality assessment system, mapping multi-source uncertainties into executable confidence levels, thereby achieving multi-dimensional and accurate assessment of the spatial perception observation information of intelligent agents.

[0008] In a first aspect, the technical solution of the present invention provides a method for evaluating the quality of space-sensing observation information, comprising the following steps: Based on the measurement objectives of each dimension, the measurement information of each dimension is transformed into normalized probabilities; Based on the concept of entropy in information theory, the normalized probability of each dimension is transformed into a computable entropy value; A downstream task weight coefficient is introduced into the entropy value of each dimension. The multidimensional entropy value after introducing the downstream task weight coefficient is mapped to the physical interpretability confidence level in the interval [0,1] through an exponential decay function. This confidence level explicitly represents the uncertainty of the observation information in each dimension.

[0009] In an optional implementation, for the spatial consistency dimension, the metric information is transformed into normalized probabilities, specifically including: definition For the Observation point in the three-dimensional space of the scene at any moment Corresponding image segmentation mask The centroid of the observation point is calculated using the following formula. The coordinates of the corresponding point on the two-dimensional image distance European distance ,

[0010] The observation points under the spatial consistency dimension are calculated using the following formula. normalized probability , .

[0011] In an optional implementation, for the geometric consistency dimension of the lidar point cloud, the metric information is transformed into a normalized probability, specifically including: Record No. Observation point in the three-dimensional space of the scene at any moment The corresponding lidar point cloud coordinates are The point cloud coordinates are calculated using the following formula. Mahalanobis distance between local point cloud distribution ,

[0012] In the formula, and They represent the observation points respectively. Local point cloud location mean and covariance matrix; The following formula is used to calculate the observation points in the geometric consistency dimension of the lidar point cloud. normalized probability , .

[0013] In an optional implementation, for the visual point cloud geometric consistency dimension, the metric information is transformed into normalized probabilities, specifically including: Combining the depth images acquired by the camera with the camera's intrinsic parameters, the first Observation point in the three-dimensional space of the scene at any moment The coordinates of the corresponding point on the two-dimensional image Perform a 2D to 3D projection to obtain visual point cloud coordinates. ; The visual point cloud coordinates are calculated using the following formula. Mahalanobis distance between the local visual point cloud distribution and the local visual point cloud distribution ,

[0014] In the formula, and They represent the observation points respectively. Local visual point cloud location mean and covariance matrix; The following formula is used to calculate the observation points in the geometric consistency dimension of the visual point cloud. normalized probability , .

[0015] In an optional implementation, for the time stability dimension, the metric information is transformed into a normalized probability, specifically including: Using Kalman filtering to track and obtain the sliding time window The first in Observation point in the three-dimensional space of the scene at any moment Position sequence on a two-dimensional image ; The observation point is calculated using the following formula. inter-frame variation evaluation value ,

[0016] The observation points under the time stability dimension are calculated using the following formula. normalized probability , .

[0017] In an optional implementation, based on the concept of entropy in information theory, the normalized probability of each dimension is transformed into a computable entropy value, specifically by calculating the entropy value using the following formula:

[0018] In the formula, For the The first dimension Observation point in the three-dimensional space of the scene at any moment The normalized entropy value, For the Observation points in dimensionality The normalized probability.

[0019] In an optional implementation, the multidimensional entropy value after introducing downstream task weight coefficients is mapped to a physical interpretability confidence level in the [0,1] interval using an exponential decay function. Specifically, this includes calculating the first... Observation point in the three-dimensional space of the scene at any moment confidence level ,

[0020] In the formula, For the Downstream task weight coefficients under the dimension For the Observation points in dimensionality The normalized entropy value.

[0021] Secondly, the technical solution of the present invention provides a space sensing observation information quality assessment system, comprising: The metric probability transformation module is used to transform the metric information of each dimension into normalized probabilities based on the metric target of each dimension. The metric entropy value conversion module is used to convert the normalized probability of each dimension into a computable entropy value based on the concept of entropy in information theory. The confidence calculation module is used to introduce downstream task weight coefficients into the entropy value of each dimension. The multidimensional entropy value after introducing the downstream task weight coefficients is mapped to the physical interpretability confidence value in the interval [0,1] through an exponential decay function. This confidence value explicitly represents the uncertainty of the observation information in each dimension.

[0022] Thirdly, the technical solution of the present invention provides a terminal, comprising: Memory, used to store the quality assessment program for space-sensing observation information; A processor is configured to implement the steps of the space-aware observation information quality assessment method as described above when executing the space-aware observation information quality assessment program.

[0023] Fourthly, the present invention provides a computer-readable storage medium storing a space-sensing observation information quality assessment program, wherein the space-sensing observation information quality assessment program, when executed by a processor, implements the steps of the space-sensing observation information quality assessment method as described in any of the above claims.

[0024] As can be seen from the above technical solutions, this application has the following advantages: 1. This application transforms multi-source uncertainties (sensor noise / algorithm error / environmental disturbance) into calculable physical quantities through an entropy system, thereby achieving explicit quantification of perception reliability and solving the decision blind spot caused by traditional binary judgment or black-box probability output.

[0025] 2. The confidence mapping function of this application establishes an explicit association between the entropy dimension and physical defects (such as point cloud fractures), supports accurate fault attribution, and realizes the interpretability of fault root causes.

[0026] 3. This application can dynamically adjust the contribution rate of each entropy dimension through the downstream task weight coefficient, and the same system can seamlessly switch between multiple task modes, improving the task adaptability of the system and increasing the flexibility of scenario adaptation.

[0027] 4. This application further considers not only the spatial dimension but also the temporal dimension, realizing spatiotemporal dual-dimensional analysis, and achieving accurate assessment of the quality of intelligent agent observation information in dynamic scenarios.

[0028] 5. This application further reduces the entropy calculation to only basic geometric operations (centroid offset / Madara distance), achieving lightweight calculation and single-point evaluation latency of <5ms, thus meeting the requirements of real-time robot control. Attached Figure Description

[0029] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a schematic diagram of a method for assessing the quality of space-sensing observation information provided in an embodiment of the present invention.

[0031] Figure 2 This is a schematic block diagram of a space perception observation information quality assessment system provided in an embodiment of the present invention.

[0032] Figure 3 This is a schematic diagram of the structure of a terminal provided in an embodiment of the present invention. Detailed Implementation

[0033] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0035] Figure 1 This is a schematic flowchart of a method for assessing the quality of space-sensing observation information provided in an embodiment of the present invention. Figure 1 The executing entity can be a space-sensing observation information quality assessment system. The space-sensing observation information quality assessment method provided in this embodiment of the invention is executed by a computer device; correspondingly, the space-sensing observation information quality assessment system runs on the computer device. Depending on different requirements, the order of the steps in this flowchart can be changed, and some steps can be omitted.

[0036] like Figure 1 As shown, the method includes the following steps.

[0037] S1, based on the measurement objective of each dimension, transforms the measurement information of each dimension into normalized probabilities.

[0038] S2, based on the concept of entropy in information theory, transforms the normalized probability of each dimension into a computable entropy value.

[0039] S3 introduces downstream task weight coefficients into the entropy value of each dimension. The multidimensional entropy value after introducing the downstream task weight coefficients is mapped to the physical interpretability confidence level in the interval [0,1] through an exponential decay function. This confidence level explicitly represents the uncertainty of the observation information in each dimension.

[0040] The method in this embodiment establishes a complete link of multidimensional uncertainty quantification, interpretable confidence mapping, and task adaptive decision-making, thus solving the problem of perceived reliability assessment.

[0041] Furthermore, as a refinement and extension of the specific implementation methods described above, to fully illustrate the specific implementation process in this embodiment, another method for assessing the quality of spatial perception observation information is provided. This method first constructs an assessment system from multiple dimensions, including spatial consistency, geometric consistency, and temporal stability, transforming heterogeneous uncertainties such as sensor noise, environmental disturbances, and algorithm errors into calculable entropy values. Secondly, it designs an exponential decay function to map multidimensional entropy values ​​to the physically interpretable confidence level in the [0,1] interval, explicitly characterizing the uncertainty of observation information in each dimension. Finally, it introduces a task weight coefficient to dynamically adjust the contribution rate of each entropy dimension, enabling the same architecture to flexibly adapt to differentiated scenario requirements such as navigation and precision operation, achieving essential alignment between perception assessment and task objectives. It should be noted that the implementation of this invention requires an intelligent agent device with autonomous mobility. The intelligent agent device is equipped with a computing unit and sensors such as visual sensors, lidar sensors, and IMUs. The computing unit has a basic operating system and is equipped with fundamental tools for image processing (including image recognition algorithms such as YOLO and task- and scene-related model training) and point cloud processing, as well as the technical environment required for downstream tasks such as arbitrary independent LiDAR mapping and localization or multi-sensor fusion mapping and localization. Specifically, it includes the following steps.

[0042] SS1, obtain the first Time observation point LiDAR point cloud coordinates The coordinates of the corresponding point on the two-dimensional image and masks .

[0043] The system utilizes sensors such as visual sensors, lidar sensors, and IMUs on the intelligent device itself to collect and preprocess observation data, primarily focusing on quality assessment of the observation data from visual sensors and lidar sensors. For the first... time( ), observation points in the three-dimensional space of the scene The corresponding lidar point cloud coordinates are Using lidar point cloud projection technology, observation points in three-dimensional space are projected... Projected to the Images are captured in real time. Above, obtain The coordinates of the corresponding point on the two-dimensional image For the first Images are captured in real time. YOLOv8 is used for segmentation to obtain a mask. Among them, observation points correspond .

[0044] SS2, Construct a confidence calculation model for spatial perception observation information.

[0045] First, based on the measurement objectives of each dimension, the measurement information of each dimension is transformed into a normalized probability. Then, based on the concept of entropy in information theory, the normalized probability of each dimension is transformed into a computable entropy value. Finally, downstream task weight coefficients are introduced into the entropy value of each dimension, and the multidimensional entropy value after introducing the downstream task weight coefficients is mapped to the physical interpretability confidence level in the interval [0,1] through an exponential decay function. This confidence level explicitly represents the uncertainty of the observation information of each dimension.

[0046] Specifically, the first is calculated according to information theory. Time observation point The confidence level is (0 = unreliable, 1 = optimal)

[0047] in, It is the first Observation points in dimensionality The normalized entropy value, For the Downstream task weight coefficients under each dimension.

[0048] According to the definition of entropy in information theory, higher uncertainty corresponds to a higher entropy value, and thus a lower confidence level. The normalized entropy value is derived from the normalized probability and is expressed as:

[0049] in, For the Observation points in dimensionality The normalized probability is obtained by converting the measurement information of each dimension into a normalized probability based on the measurement target of each dimension.

[0050] (1) Spatial consistency dimension Let the spatial consistency dimension be denoted as the first dimension. Below the first dimension, the entropy value is calculated. Measuring three-dimensional spatial observation points Projection on a two-dimensional image Relative to its corresponding image segmentation mask The degree of deviation. The greater the deviation, the lower the reliability of the observation point.

[0051] For the spatial consistency dimension, the measurement information is transformed into normalized probabilities, which specifically includes the following steps.

[0052] Step 1, Define For observation point Corresponding image segmentation mask The centroid is calculated. distance European distance as follows,

[0053] Step 2, calculate the observation points under the spatial consistency dimension. normalized probability , is represented as ,

[0054] (2) Geometric consistency dimension of lidar point cloud The geometric consistency dimension of the lidar point cloud is denoted as the second dimension. For lidar observation data, entropy is used... Describe the three-dimensional space observation point Corresponding lidar point cloud coordinates The degree of deviation from the local laser point cloud. The greater the deviation, the lower the reliability of the observation point.

[0055] For the geometric consistency dimension of LiDAR point clouds, the metric information is transformed into normalized probabilities, which specifically includes the following steps.

[0056] Step 1: Calculate the point cloud coordinates using the following formula. Mahalanobis distance between local point cloud distribution ,

[0057] In the formula, and They represent the observation points respectively. Mean and covariance matrix of local point cloud locations.

[0058] Step 2: Calculate the observation points under the geometric consistency dimension of the lidar point cloud. normalized probability , is represented as ,

[0059] (3) Visual point cloud geometric consistency dimension Let the geometric consistency dimension of the visual point cloud be denoted as the third dimension, for three-dimensional spatial observation points. The coordinates of the corresponding point on the two-dimensional image By combining depth images captured by an RGB-D camera with camera intrinsic parameters, a 2D-to-3D projection is performed to obtain visual point cloud coordinates. To evaluate the quality of depth values ​​in data acquired by an RGB-D camera, entropy values ​​are used for visual point clouds calculated based on depth maps. Describe the three-dimensional space observation point Corresponding visual point cloud coordinates The degree of deviation from the local visual point cloud. Three-dimensional spatial observation point. The corresponding local visual point cloud is formed by image segmentation mask. This was calculated in conjunction with the depth map. Large deviations point to outliers in the depth observation information from the RGB-D camera.

[0060] For the geometric consistency dimension of visual point clouds, the metric information is transformed into normalized probabilities, which specifically includes the following steps.

[0061] Step 1: Calculate the visual point cloud coordinates using the following formula. Mahalanobis distance between the local visual point cloud distribution and the local visual point cloud distribution ,

[0062] In the formula, and They represent the observation points respectively. Mean and covariance matrix of local visual point cloud locations.

[0063] Step 2: Calculate the observation points under the geometric consistency dimension of the visual point cloud. normalized probability , is represented as ,

[0064] (4) Time stability dimension By designating temporal stability as the fourth dimension, and further considering the temporal dimension, we can evaluate the stability of spatially sensed observation information under dynamic environments. Therefore, we calculate the entropy value. To measure the observation points in three-dimensional space The degree of change over a certain time span.

[0065] For the time stability dimension, the measurement information is transformed into normalized probabilities, which specifically includes the following steps.

[0066] Step 1: Use Kalman filtering to track and obtain the sliding time window. The first in Observation point in the three-dimensional space of the scene at any moment Position sequence on a two-dimensional image .

[0067] Step 2, Calculate the observation points inter-frame variation evaluation value , is represented as ,

[0068] Step 3: Calculate the observation points under the time stability dimension. normalized probability , is represented as ,

[0069] Downstream task weight coefficient These are weighting factors adjusted based on downstream tasks. Through confidence level calculations, high-quality observation information is selected for use in downstream tasks crucial to the agent, such as mapping and localization, navigation, and robotic arm operations.

[0070] In some alternative implementations, the weighting factor is adjusted based on the characteristics of the downstream task and the scenario. The following adjustments can be made:

[0071] The above text provides a detailed description of an embodiment of a method for assessing the quality of space-sensing observation information. Based on the space-sensing observation information quality assessment method described in the above embodiment, this invention also provides a space-sensing observation information quality assessment system corresponding to the method.

[0072] Figure 2 This is a schematic block diagram of a space-sensing observation information quality assessment system provided in an embodiment of the present invention. In this embodiment, the space-sensing observation information quality assessment system 200 can be divided into multiple functional modules according to the functions it performs. A module, as referred to in this invention, is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and is stored in memory.

[0073] The metric probability transformation module 210 is used to transform the metric information of each dimension into normalized probabilities based on the metric objective of each dimension.

[0074] The entropy value conversion module 220 is used to convert the normalized probability of each dimension into a computable entropy value based on the concept of entropy in information theory.

[0075] The confidence calculation module 230 is used to introduce downstream task weight coefficients into the entropy value of each dimension. The multidimensional entropy value after introducing the downstream task weight coefficients is mapped to the physical interpretability confidence in the interval [0,1] through an exponential decay function. This confidence explicitly characterizes the uncertainty of the observation information in each dimension.

[0076] The space perception observation information quality assessment system of this embodiment is used to implement the aforementioned space perception observation information quality assessment method. Therefore, the specific implementation of this system can be found in the embodiment section of the space perception observation information quality assessment method above. Thus, the specific implementation can be referred to the description of the corresponding embodiments, and will not be elaborated here.

[0077] Furthermore, since the spatial sensing observation information quality assessment system of this embodiment is used to implement the aforementioned spatial sensing observation information quality assessment method, its function corresponds to the function of the above method, and will not be repeated here.

[0078] Figure 3 A schematic diagram of a terminal 300 provided in an embodiment of the present invention includes: a processor 310, a memory 320, and a communication unit 330. The processor 310 is used to implement the following steps when implementing the space sensing observation information quality assessment program stored in the memory 320: Based on the measurement objectives of each dimension, the measurement information of each dimension is transformed into normalized probabilities; Based on the concept of entropy in information theory, the normalized probability of each dimension is transformed into a computable entropy value; A downstream task weight coefficient is introduced into the entropy value of each dimension. The multidimensional entropy value after introducing the downstream task weight coefficient is mapped to the physical interpretability confidence level in the interval [0,1] through an exponential decay function. This confidence level explicitly represents the uncertainty of the observation information in each dimension.

[0079] The terminal 300 includes a processor 310, a memory 320, and a communication unit 330. These components communicate via one or more buses. Those skilled in the art will understand that the server structure shown in the figure does not constitute a limitation of the present invention. It can be a bus topology or a star topology, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0080] The memory 320 can be used to store the execution instructions of the processor 310. The memory 320 can be implemented by any type of volatile or non-volatile memory terminal or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. When the execution instructions in the memory 320 are executed by the processor 310, the terminal 300 is able to perform some or all of the steps in the above method embodiments.

[0081] The processor 310 serves as the control center of the storage terminal, connecting various parts of the electronic terminal via various interfaces and lines. It executes software programs and / or modules stored in the memory 320, and calls data stored in the memory to perform various functions of the electronic terminal and / or process data. The processor can be composed of integrated circuits (ICs), such as a single packaged IC or multiple packaged ICs with the same or different functions connected together. For example, the processor 310 may consist only of a central processing unit (CPU). In this embodiment of the invention, the CPU may have a single processing core or include multiple processing cores.

[0082] The communication unit 330 is used to establish a communication channel, enabling the storage terminal to communicate with other terminals. It can receive user data sent by other terminals or send user data to other terminals.

[0083] The present invention also provides a computer storage medium, which may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0084] The present invention also provides a computer storage medium, which may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0085] The computer storage medium stores a space-sensing observation information quality assessment program. When the space-sensing observation information quality assessment program is executed by the processor, it performs the following steps: Based on the measurement objectives of each dimension, the measurement information of each dimension is transformed into normalized probabilities; Based on the concept of entropy in information theory, the normalized probability of each dimension is transformed into a computable entropy value; A downstream task weight coefficient is introduced into the entropy value of each dimension. The multidimensional entropy value after introducing the downstream task weight coefficient is mapped to the physical interpretability confidence level in the interval [0,1] through an exponential decay function. This confidence level explicitly represents the uncertainty of the observation information in each dimension.

[0086] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium such as a USB flash drive, mobile hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, or other media capable of storing program code. It includes several instructions to cause a computer terminal (which may be a personal computer, server, or a second terminal, network terminal, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0087] In the embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0088] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0089] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0090] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for assessing the quality of space-sensing observation information, characterized in that, Includes the following steps: Based on the measurement objectives of each dimension, the measurement information of each dimension is transformed into normalized probabilities; Based on the concept of entropy in information theory, the normalized probability of each dimension is transformed into a computable entropy value; A downstream task weight coefficient is introduced into the entropy value of each dimension. The multidimensional entropy value after introducing the downstream task weight coefficient is mapped to the physical interpretability confidence level in the interval [0,1] through an exponential decay function. This confidence level explicitly represents the uncertainty of the observation information in each dimension.

2. The method for assessing the quality of space-sensing observation information according to claim 1, characterized in that, For the spatial consistency dimension, the metric information is transformed into normalized probabilities, specifically including: definition For the Observation point in the three-dimensional space of the scene at any moment Corresponding image segmentation mask The centroid of the observation point is calculated using the following formula. The coordinates of the corresponding point on the two-dimensional image distance European distance , The observation points under the spatial consistency dimension are calculated using the following formula. normalized probability , 。 3. The method for assessing the quality of space-sensing observation information according to claim 1, characterized in that, For the geometric consistency dimension of LiDAR point clouds, the metric information is transformed into normalized probabilities, specifically including: Record No. Observation point in the three-dimensional space of the scene at any moment The corresponding lidar point cloud coordinates are The point cloud coordinates are calculated using the following formula. Mahalanobis distance between local point cloud distribution , In the formula, and They represent the observation points respectively. Local point cloud location mean and covariance matrix; The following formula is used to calculate the observation points in the geometric consistency dimension of the lidar point cloud. normalized probability , 。 4. The method for assessing the quality of space-sensing observation information according to claim 1, characterized in that, For the geometric consistency dimension of visual point clouds, the metric information is transformed into normalized probabilities, specifically including: Combining the depth images acquired by the camera with the camera's intrinsic parameters, the first Observation point in the three-dimensional space of the scene at any moment The coordinates of the corresponding point on the two-dimensional image Perform a 2D to 3D projection to obtain visual point cloud coordinates. ; The visual point cloud coordinates are calculated using the following formula. Mahalanobis distance between the local visual point cloud distribution and the local visual point cloud distribution , In the formula, and They represent the observation points respectively. Local visual point cloud location mean and covariance matrix; The following formula is used to calculate the observation points in the geometric consistency dimension of the visual point cloud. normalized probability , 。 5. The method for assessing the quality of space-sensing observation information according to claim 1, characterized in that, For the time stability dimension, the metric information is transformed into normalized probabilities, specifically including: Using Kalman filtering to track and obtain the sliding time window The first in Observation point in the three-dimensional space of the scene at any moment Position sequence on a two-dimensional image ; The observation point is calculated using the following formula. inter-frame variation evaluation value , The observation points under the time stability dimension are calculated using the following formula. normalized probability , 。 6. The method for assessing the quality of space-sensing observation information according to claim 1, characterized in that, Based on the concept of entropy in information theory, the normalized probability of each dimension is transformed into a computable entropy value, specifically by calculating the entropy value using the following formula. In the formula, For the The first dimension Observation point in the three-dimensional space of the scene at any moment The normalized entropy value, For the Observation points in dimensionality The normalized probability.

7. The method for assessing the quality of space-sensing observation information according to claim 1, characterized in that, The multidimensional entropy value, after introducing the weight coefficients of downstream tasks, is mapped to the physical interpretability confidence value in the interval [0,1] using an exponential decay function. Specifically, this includes calculating the first... Observation point in the three-dimensional space of the scene at any moment confidence level , In the formula, For the Downstream task weight coefficients under the dimension For the Observation points in dimensionality The normalized entropy value.

8. A space-sensing observation information quality assessment system, characterized in that, include: The metric probability transformation module is used to transform the metric information of each dimension into normalized probabilities based on the metric target of each dimension. The metric entropy value conversion module is used to convert the normalized probability of each dimension into a computable entropy value based on the concept of entropy in information theory. The confidence calculation module is used to introduce downstream task weight coefficients into the entropy value of each dimension. The multidimensional entropy value after introducing the downstream task weight coefficients is mapped to the physical interpretability confidence value in the interval [0,1] through an exponential decay function. This confidence value explicitly represents the uncertainty of the observation information in each dimension.

9. A terminal, characterized in that, include: Memory, used to store the quality assessment program for space-sensing observation information; A processor is configured to implement the steps of the space-aware observation information quality assessment method as described in any one of claims 1 to 7 when executing the space-aware observation information quality assessment program.

10. A computer-readable storage medium, characterized in that, The readable storage medium stores a space-aware observation information quality assessment program, which, when executed by a processor, implements the steps of the space-aware observation information quality assessment method as described in any one of claims 1 to 7.

Images