A method for extracting meteor points from dynamic events based on Bézier curves
By employing a dynamic event meteor point extraction method based on Bézier curves, and utilizing incremental clustering and adaptive noise removal, the problem of star point extraction accuracy in high dynamic environments is solved, achieving high-precision and high-dynamic-performance star point detection.
Patent Information
- Application Number
- CN202311470267.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-07
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-11-07
AI Technical Summary
Existing technologies struggle to achieve high-precision extraction of stellar points in highly dynamic environments. Traditional methods suffer from motion blur and fail to fully utilize the advantages of dynamic vision sensors.
A dynamic event meteor point extraction method based on Bézier curves is adopted. Initial detection is performed through incremental clustering and adaptive noise removal, Bézier curves are established for accurate continuous position estimation, and an adaptive meteor point detection and update strategy is combined to achieve high-precision extraction.
It improves the accuracy and dynamic performance of star point extraction in highly dynamic environments, and achieves high-precision star point detection.
Smart Images

Figure CN117421613B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of high dynamic stellar point extraction technology, and in particular to a method for extracting dynamic event meteor points based on Bézier curves. Background Technology
[0002] As space missions place increasingly higher demands on the precision of spacecraft attitude control and enhance their dynamic performance, high attitude measurement accuracy remains crucial under extremely dynamic environments, particularly during high-speed maneuvers and rapid orbit changes. While star sensors currently offer the highest attitude measurement accuracy for spacecraft, their limitations imposed by traditional image sensor principles make it difficult to achieve highly dynamic attitude measurement while still meeting detection requirements. Therefore, achieving high-precision attitude measurement hinges on the precise extraction of the centroid of star points.
[0003] Traditional image sensors, in principle, achieve image formation of a target through a certain exposure time. However, in high dynamic scenes, traditional image sensors suffer from motion blur, making it difficult to achieve high-precision centroid extraction of star points in high dynamic scenes. Therefore, star sensors developed based on traditional image sensors are inherently difficult to achieve high-precision centroid extraction in high dynamic scenes, thus making it difficult to achieve high-precision attitude measurement.
[0004] Current star point extraction algorithms based on dynamic vision sensors (event cameras) mainly extract star points by converting flow event data into two-dimensional image data, failing to effectively utilize the advantages of high dynamics, high temporal resolution, and low latency of dynamic vision sensors.
[0005] Both patent applications with patent numbers 202211380360.X and 202210142077.7 calculate centroid coordinates by integrating event streams to form event frames and inheriting the traditional image star point centroid extraction algorithm. This improves dynamic performance to some extent, but fails to solve the motion blur problem in high dynamic environments.
[0006] The patent application with patent number 202210414348.X uses a time-slicing method, which accumulates event stream data over a certain period of time to obtain a two-dimensional image containing star points, thereby realizing the extraction of star points. However, it also relies on the concept of event frames, failing to fully leverage the advantages of dynamic event sensing.
[0007] Existing technologies also include methods for extracting star points, which involve accumulating and integrating event stream data over a certain period of time to obtain event frames, and then extracting star points based on traditional star point centroid extraction algorithms. Although star points can be extracted, the motion blur problem is not fundamentally solved, resulting in low extraction accuracy.
[0008] None of the above methods involve the dynamic event stream star point extraction method based on Bézier curves proposed in this invention. In summary, existing star point extraction methods based on dynamic event streams suffer from problems such as reliance on traditional centroid extraction methods, low extraction accuracy, and failure to fully utilize dynamic visual sensing modes. Summary of the Invention
[0009] To address this, the present invention proposes a dynamic event meteor point extraction method for Bézier curves. This method does not perform dimensionality reduction processing on event stream data, does not rely on traditional centroid extraction methods, can fully leverage the advantages of dynamic visual sensing, and has advantages such as high extraction accuracy, strong adaptability to dynamic environments, and continuous estimation.
[0010] The technical problem solved by this invention is to provide a high-precision extraction method for stellar constellations under highly dynamic environments. This method utilizes incremental clustering and adaptive noise removal of dynamic event stream data to achieve initial detection of stellar constellations. Then, it establishes Bézier curves on the event stream data constituting the constellations to accurately estimate their continuous positions. Simultaneously, it combines an adaptive constellation detection and update strategy to achieve accurate extraction of stellar constellations under highly dynamic motion environments, effectively improving dynamic performance.
[0011] The technical solution of this invention is as follows:
[0012] (1) The event camera continuously images the starry sky to obtain the event stream sequence E. n ;
[0013] (2) Initialize star point detection. Initialize the event stream sequence obtained in step (1) by clustering star point events to construct a star point list {P}. k =(u k ,v k The initialization of star point detection includes two steps: incremental clustering and adaptive noise removal. First, incremental clustering is performed on the event stream sequence obtained in step (1), that is, cluster analysis is performed on each event. The events are classified by drift category centers. The cumulative time T0 is used as the threshold for the end of initialization, thus obtaining the star point list L. u Simultaneously, adaptive noise removal is performed during cluster analysis; that is, within the Δτ event, clustering events are less than the threshold Th. τ This category is then identified as noise and removed, finally resulting in the initial star point list {P}. k =(u k ,v k )};P k To initialize the star point, u k Let v be the coordinates of the star point. k The coordinates of the star point;
[0014] (3) Construct the Bézier curve, based on the star point list L obtained in step (2).u For each star point containing event data, construct a Bézier curve according to the following formula:
[0015]
[0016] Where n is the degree of the Bézier curve, p is the selected control point, and t is the curve coefficient taking the value [0,1]. It is a Bernstein basis polynomial, p n (t) represents the Bézier curve;
[0017] Based on the Bézier curves constructed using the above formula for multiple star points, the star point coordinates in the event stream information are aligned along the time axis to the Bézier curves. Simultaneously, a cubic Bézier curve fitting is performed on the time axis. n This allows for a continuous estimation of star points over time.
[0018] (4) Estimate the position of the star points in real time, based on the Bézier curves B of all star points within the field of view obtained in step (3). n B n The curve is a function with time as the variable, and the position S of the star point can be calculated and updated in real time based on the current moment. n ;
[0019] (5) Adaptive star point detection and update, based on the dynamically updated star point position S in step (4). n Dynamically calculate the event stream data and the current star point position S. n Given a distance d, determine whether the threshold Th is met. d Classify events:
[0020]
[0021] For the star points to which newly added event data belongs, update and calculate the extended Bézier curve cubic fitting function B. n .
[0022] When a new event does not belong to any star point, incremental clustering is performed on T. a The star detection threshold Th is met within the time limit. s The update is for newly entering stars in the field of view, while also updating T. out Stars that do not have new events satisfying the distance threshold within a given time period are removed, thus enabling the management of the star list L. u Adaptive detection and update.
[0023] The present invention has the following beneficial technical results:
[0024] This paper presents a method for extracting star points based on dynamic event streams. The method utilizes incremental clustering to analyze the category of each event while adaptively removing noisy categories. Bézier curves are constructed based on the star event clusters obtained from the classification, and star positions are estimated in real time based on these curves. Finally, the star list is adaptively updated based on the star positions and the Bézier curves of each star, thus achieving high-precision and highly dynamic star point extraction. Attached Figure Description
[0025] Figure 1 This is a flowchart illustrating the specific process of this invention;
[0026] Figure 2 This is a flowchart of the initial star detection process for this invention;
[0027] Figure 3 This is a flowchart of the adaptive star point detection and update process of the present invention;
[0028] Figure 4 This is a diagram showing the error in star point extraction according to the present invention. Detailed Implementation
[0029] The method of the present invention will be further described below with reference to specific embodiments.
[0030] The specific process of this invention is as follows: Figure 1 As shown.
[0031] (1) Continuously image the starry sky and acquire a 10-second dynamic event stream sequence E. n The event stream sequence contains 8 star points;
[0032] (2) Initialize star point detection. Initialize the event stream sequence obtained in step (1) by clustering star point events to construct a star point list {P}. k =(u k ,v k Specifically:
[0033] (2-1) Incremental clustering to complete the initial detection of star points. First, the event stream E obtained in step (1) is processed. n Incremental clustering is performed event by event. First, the first event e0 is used as the center, and the flow events are classified one by one. The cumulative time T0 is used as the threshold for the initialization end judgment.
[0034] (2-2) Adaptive noise removal: Noise removal is performed simultaneously with step (2-1). For clustering events, within the time interval Δτ, if the clustering event is less than the threshold Th... τ This category is then identified as noise and removed, finally resulting in the initial star point list {P}. k =(u k ,v k )}。 P kTo initialize the star point, u k Let v be the coordinates of the star point. k The initialization process for the star point's row coordinates is as follows: Figure 2 As shown;
[0035] (3) Construct the Bézier curve, based on the star point list L obtained in step (2). u For each star point containing event data, construct a Bézier curve according to the following formula:
[0036]
[0037] Where n is the degree of the Bézier curve, p is the selected control point, and t is the curve coefficient taking the value [0,1]. It is a Bernstein basis polynomial.
[0038] Based on the Bézier curves constructed using the above formula for multiple star points, the star point coordinates in the event stream information are aligned along the time axis to the Bézier curves. Simultaneously, a cubic fit of the extended Bézier curves is performed on the time axis according to the following formula: n This allows for a continuous estimation of star points over time.
[0039]
[0040] (4) Estimate the position of the star points in real time, based on the Bézier curves B of all star points within the field of view obtained in step (3). n B n The curve is a function with time as the variable, and the position S of the star point can be calculated and updated in real time based on the current moment. n :
[0041] (5) Adaptive star point detection and update, based on the dynamically updated star point position S in step (4). n Dynamically calculate the event stream data and the current star point position S. n Given a distance d, determine whether the threshold Th is met. d The events are categorized, and the process is as follows: Figure 3 As shown;
[0042]
[0043] For the star points to which newly added event data belongs, update and calculate the extended Bézier curve cubic fitting function B. n .
[0044] When a new event does not belong to any star point, incremental clustering is performed on T. a The star detection threshold Th is met within the time limit. s The update is for newly entering stars in the field of view, while also updating T. outStars that do not have new events satisfying the distance threshold within a given time period are removed, thus enabling the management of the star list L. u Adaptive detection and update, star point extraction error such as Figure 4 As shown.
[0045] The parts of this invention not described in detail belong to the well-known technology in this field. The above description is only a specific example of this invention and is not intended to limit this invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the protection scope of this invention.
Claims
1. A method for extracting meteor points from dynamic events based on Bézier curves, characterized in that, The method includes the following steps: Step (1): The event camera continuously images the starry sky to obtain the event stream sequence E. n ; Step (2): Initialize star point detection; initialize the star point event clustering of the event stream sequence obtained in step (1) to construct a star point list {P}. k =(u k ,v k The initialization of star point detection includes two steps: incremental clustering and adaptive noise removal. First, incremental clustering is performed on the event stream sequence obtained in step (1), that is, cluster analysis is performed on each event. The events are classified by drift category centers. The cumulative time T0 is used as the threshold for the end of initialization, thus obtaining the star point list L. u Simultaneously, adaptive noise removal is performed during cluster analysis; that is, within the Δτ event, clustering events are less than the threshold Th. τ This category is then identified as noise and removed, finally resulting in the initial star point list {P}. k =(u k ,v k )}, P k To initialize the star point, u k Let v be the column coordinates of the star point. k The coordinates of the star point; Step (3): Construct the Bézier curve, based on the star point list L obtained in step (2). u For each star point containing event data, construct a Bézier curve according to the following formula: Where n is the degree of the Bézier curve, p i Here are the selected control points, and t is the curve coefficient taking the value [0,1]. It is a Bernstein basis polynomial, p n (t) represents the Bézier curve; Based on the Bezier curves of multiple star points constructed in the above formula, the coordinates of star points in the event stream information are aligned to the Bezier curves along the time axis. At the same time, extended Bezier curves are fitted cubically on the time axis to make the star points continuously estimated in the time dimension. Step (4): Estimate the position of the star points in real time, based on the Bézier curves B of all star points within the field of view obtained in step (3). n B n The curve is a function with time as the variable, which can calculate and update the position of the star points in real time based on the current moment; Step (5): Adaptive star point detection and update, based on the star point position S dynamically updated in step (4). n Dynamically calculate the event stream data and the current star point position S. n Given a distance d, determine whether the threshold Th is met. d Classify events: For the star points to which newly added event data belongs, update and calculate the extended Bézier curve cubic fitting function B. n ; When a new event does not belong to any star point, incremental clustering is performed on T. a The star detection threshold Th is met within the time limit. s The update is for newly entering stars in the field of view, while also updating T. out Stars that do not have new events satisfying the distance threshold within a given time period are removed, thus enabling the management of the star list L. u Adaptive detection and update.
2. The method for extracting dynamic event meteor points based on Bézier curves according to claim 1, characterized in that: The incremental clustering initialization of the star list proposed in step (2) should specifically be completed by two steps: incremental clustering and adaptive noise removal. Incremental clustering completes the classification of all events within time T0, and adaptive noise removal completes the full removal of noise events, thereby achieving the initialization and establishment of the star list.
3. The method for extracting dynamic event meteor points based on Bézier curves according to claim 1, characterized in that: The construction of the Bézier curve in step (3) is based on the star point list obtained in step (2), and the event is classified according to the star point to which it belongs, according to the formula. Calculate the Bézier curve.
4. The method for extracting dynamic event meteor points based on Bézier curves according to claim 1, characterized in that: The adaptive star point detection and update method proposed in step (5) classifies the dynamic event stream according to the star point classification to which the event belongs, and performs incremental clustering on events that do not belong to the star point list at intervals ΔT, thereby dynamically updating the star point list.
Citation Information
Patent Citations
A method for determining the attitude of an aircraft using an event camera as a star sensor
CN114663492B
A method for star point extraction using a high dynamic range star sensor based on an event camera
CN114777764B
A star map recognition method based on an event camera
CN115900728B
Method for solving parameters of radiation distribution of equivalent Gaussian beam far-field
CN102622511A
Construction method for hyperboloid furred ceiling
CN105133840A
