Event camera seeker system for high-speed spinning carrier

Abstract

The invention relates to an event camera seeker system for a high-speed spin carrier. Comprising an event camera sensor which is used for carrying out asynchronous perception on a scene under a carrier spin motion condition, generating and outputting event stream data containing pixel coordinates, timestamps and polarities, and recording event streams as an event set; the inertial measurement module is fixedly mounted in the carrier, outputs attitude parameters, and realizes synchronization with an event timestamp through a time reference or an alignment circuit shared by the inertial measurement module and the event camera sensor; and the event flow target detection module is used for performing effective event screening and grouping, motion compensation, dynamic and static separation, target matching and positioning and target azimuth information calculation operations on the event set according to a preset time sequence after receiving the event set and the attitude parameters, outputting target azimuth information and outputting the target information to a guide control system. According to the invention, stable and real-time target azimuth output can be realized under high-speed spinning and high-dynamic conditions.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of event vision and guidance control, and particularly relates to an event camera seeker system for a high-speed spinning carrier. BACKGROUND

[0002] Precise guidance is the core capability of modern ammunition, and the performance of the terminal seeker directly determines the strike effectiveness. Visible light cameras are widely used in guidance systems due to their low cost and rich information. However, the projectile body is often spun at high speed to maintain flight stability, which causes severe motion blur and geometric distortion in the imaging of traditional frame cameras, significantly affecting target detection and recognition accuracy. In the prior art, the gimbal stabilization scheme has a complex structure and poor overload resistance, and the strapdown scheme has extremely high requirements for the accuracy of the motion measurement sensor and computing resources.

[0003] Event cameras output event streams in an asynchronous manner, have microsecond-level time resolution, high dynamic range, and low latency, and can more completely retain target motion information in high-speed motion scenarios. However, under high-speed spinning conditions, the event stream is superimposed with global rotational motion caused by carrier spinning and relative motion of the target itself, and the foreground target events and static background events are interwoven in space, making it still challenging to extract target events and stably position them. Therefore, an event camera seeker system for a high-speed spinning carrier is needed, which realizes separation of static events and dynamic events and real-time positioning through a reasonable hardware platform and event stream processing chain, to meet the guidance control requirements. SUMMARY

[0004] The application aims to provide an event camera seeker system for a high-speed spinning carrier, to solve the problems of easy mixing of event data under superimposition of spinning and target relative motion, unstable target detection, and limited real-time closed loop in the existing high-speed spinning guidance scene, so that the seeker can continuously and real-time output target azimuth information for guidance control under strong spinning and high dynamic conditions.

[0005] The technical solution for achieving the application is an event camera seeker system for a high-speed spinning carrier, which comprises:

[0006] An event camera sensor: used for asynchronous sensing of a scene under carrier spinning motion, generating and outputting event stream data containing pixel coordinates, time stamps, and polarity, and recording the event stream as an event set ;

[0007] An inertial measurement module: fixedly installed inside the carrier, outputting attitude parameters, and realizing synchronization with event time stamps through a time reference or alignment circuit shared with the event camera sensor;

[0008] An event stream target detection module: receiving the event set After obtaining the attitude parameters, the event set is processed according to a preset timing sequence. Perform valid event filtering and grouping, motion compensation, static-dynamic separation, target matching and positioning, and target orientation information calculation operations, output target orientation information, and output the target information to the guidance and control system.

[0009] Furthermore, the event flow target detection module is located inside the seeker head compartment and includes at least:

[0010] Event stream receiving and buffering unit: used for receiving and buffering event sets. Perform continuous access, caching, and ordered output by timestamp;

[0011] Attitude parameter interface unit: used to receive attitude parameters and align them with the event time base;

[0012] Processor and storage unit: used to calculate and output target orientation information based on the event stream and aligned attitude parameters;

[0013] The target information output unit is used to output the target orientation information to the guidance and control system.

[0014] Furthermore, the multi-level buffer of the event stream receiving and caching unit includes at least an on-chip cache and an external high-speed storage cache, which are used to avoid packet loss when the event stream rate fluctuates and to keep the events output in a monotonically increasing order.

[0015] Furthermore, the time alignment circuit of the attitude parameter interface unit is used to map the attitude parameter time stamp to the event time base, so that each event group can read the attitude parameters corresponding to its time coverage interval when the processor processes it.

[0016] A method for guidance using the aforementioned seeker system includes the following steps:

[0017] Step (1): The event camera acquires raw data;

[0018] Step (2): The processor receives and reads the event set E from the event stream and cache unit, and performs redundancy suppression and effective event filtering and grouping on the events;

[0019] Step (3): The processor reads the attitude parameters from the attitude parameter interface unit, performs rotation alignment compensation on each group of event samples, and obtains the compensated event pixel coordinates; the static events and dynamic events are separated to obtain the separated event set;

[0020] Step (4): Cluster and match the separated event set to determine the corresponding target and output the location of the matched target;

[0021] Step (5) Based on the matched target position, calculate the horizontal deflection angle and pitch deflection angle of the target relative to the optical axis to obtain the target azimuth information, and output it to the guidance and control system in real time at a preset update cycle to form a closed loop.

[0022] Furthermore, the effective event filtering grouping in step (2) specifically involves: when the event polarity... An event is considered valid if it belongs to the positive polarity set, and the positive polarity set is... and at fixed time intervals Grouping the valid event set yields a grouped event set. ;

[0023] ,in It is a non-negative integer.

[0024] Furthermore, the rotation alignment compensation in step (3) specifically includes:

[0025] Extract the spin angular velocity vector from the attitude parameters. Regarding the event The pixel coordinates are Calculate the relative reference initial time Time difference And let the rotation angle be:

[0026] ,

[0027] in for The length of the mold, The unit rotation axis vector;

[0028] Construct the rotation correction matrix:

[0029]

[0030] in It is the identity matrix. Unit rotation axis The antisymmetric matrix;

[0031] Use event camera calibration parameters to set pixel coordinates Mapped to 3D points in the camera coordinate system and according to Three-dimensional points after compensation Then, the compensated 3D points Back projection onto the pixel plane yields the compensated pixel coordinates. .

[0032] Furthermore, the separation of static and dynamic events in step (3) specifically involves:

[0033] Within the compensation domain, the spatial consistency criterion is used for discrimination, and the initial time is determined. The event pixel coordinates are as And the compensated pixel coordinates and Calculate the Euclidean distance if it satisfies If it is a static event, it is considered a dynamic event; otherwise, it is considered a dynamic event. This is a preset pixel distance threshold.

[0034] Furthermore, step (4) specifically involves:

[0035] Compensated pixel coordinates corresponding to the separated events Density clustering is used to form event clusters. Calculate the centroid pixel coordinates of the event cluster. ,satisfy:

[0036]

[0037] The centroid pixel coordinates and event cluster scale features are then matched with the target prior information using a nearest neighbor threshold to determine the target event cluster and output the target image plane position.

[0038] Furthermore, step (5) specifically involves:

[0039] Based on the target image plane position By combining the calibration parameters of the event camera, the target image plane position is converted into the target line of sight, and the horizontal deflection angle of the target relative to the optical axis is calculated from the line of sight. With pitch and yaw angle This serves as the output of target location information.

[0040] Compared with the prior art, the significant advantages of this invention are:

[0041] (1) The event stream target detection module is composed of an event stream receiving and caching unit, an attitude parameter interface unit, a processor and storage unit and a target information output unit. It can realize high-speed access, synchronous caching and calculation of event streams, and ensure that event data under high-speed spin continuously and stably enters the processing chain, thereby reducing the impact of redundancy and loss on subsequent extraction from the hardware level.

[0042] (2) Relying on the cache and computing power of the above hardware, the event stream receiving and caching unit runs a filtering grouping function that uses effective event filtering grouping and spatiotemporal grouping to suppress and effectively filter event stream redundancy, reduce the impact of motion tailing and remove redundant data.

[0043] (3) The attitude parameters output by the inertial measurement module are synchronously input into the processor via the interface. The event flow target detection module performs rotation alignment compensation based on the spin angular velocity component in the attitude parameters, so that the static background and dynamic foreground exhibit separable characteristics in the compensation space, reducing the coupling between the background and the target event.

[0044] (4) Based on spatial consistency, static background and dynamic foreground are separated, and stable target event clusters are obtained on the processor through density clustering and prior matching. The target orientation information is further calculated and output in real time. The results are fed back to the output unit for guidance control. The system outputs continuous and reliable data in complex spin scenarios and the hardware is easy to miniaturize and integrate. Attached Figure Description

[0045] Fig. 1 This is a schematic diagram of the overall framework of the event camera seeker system for high-speed spin carriers according to the present invention. The figure shows the composition and connection relationship of the event camera sensor, the inertial measurement module, and the event flow target detection module.

[0046] Fig. 2 This is a hardware system framework diagram of the event stream target detection module of the present invention. The diagram shows the composition and data connection relationship of the event stream receiving and buffering unit, the attitude parameter interface unit, the processor and storage unit, and the target information output unit.

[0047] Fig. 3 This is a flowchart of the processing chain of the event stream target detection module of the present invention. The diagram shows the steps performed sequentially on the processor: effective event filtering and grouping, motion compensation, static-dynamic separation, target matching and positioning, and target orientation information calculation, and output of target orientation information. Detailed Implementation

[0048] The following is combined with Figs. 1-3 The illustrated embodiment provides a detailed description of the present invention. The event camera seeker system for high-speed spin carriers described in this embodiment can be deployed in scenarios such as high-speed spin missiles or rotating platforms. The system has a compact structure, and each hardware unit within the event flow target detection module carries the corresponding event processing function under a unified clock cycle, so as to stably output target azimuth information under strong spin and high dynamic conditions.

[0049] like Fig. 1 As shown, the seeker system in this embodiment comprises three main parts: an event camera sensor, an inertial measurement module, and an event flow target detection module. The event camera sensor is installed in the field of view of the seeker on the high-speed spinning carrier, and is used to asynchronously sense the external scene and output a set of events under the conditions of the carrier's spin motion. The inertial measurement module is fixed inside the carrier and is used to measure and output the carrier's attitude parameters, including the spin angular velocity vector characterizing the carrier's spin motion state. The event stream target detection module is electrically connected to the two modules mentioned above and is used to receive event sets. It outputs the target's azimuth information online along with attitude parameters.

[0050] In this embodiment, the raw event stream output by the event camera sensor is an event set. A single event is denoted as: ,in For pixel coordinates, For event timestamps, Polarity. Event set. The target is continuously output to the event stream target detection module via a high-speed data interface; attitude parameters are synchronously sent to the event stream target detection module via the attitude data interface. The event stream target detection module outputs target azimuth information under real-time processing by its internal units, and transmits it to the guidance and control system via the target information output unit to form a closed loop.

[0051] The internal structure of the event stream target detection module is as follows: Fig. 2 As shown, it includes at least an event stream receiving and caching unit, an attitude parameter interface unit, a processor and storage unit, and a target information output unit. These units not only complete the access and synchronization of the physical data path, but also carry key processing functions such as filtering and grouping, rotation alignment compensation, dynamic and static separation, clustering and localization, and orientation calculation, forming an integrated online link.

[0052] (1) This unit is directly connected to the high-speed data interface of the event camera sensor, and is configured in engineering as an "input interface + buffer" structure. The input interface is responsible for parsing and receiving the event set. The cache is used to store events in order of timestamp. It also performs burst flow stabilization; during cache scheduling, this unit runs a valid event filtering and grouping function: based on event polarity Select positive polarity events to form an effective event stream, and use the initial time as the starting point. Based on a fixed time interval Group the valid events and output the first one. Group Event Set , in It is a non-negative integer. They are sequentially fed into the processor and storage unit, ensuring that the event stream under high spin continuously enters the subsequent computing link at equal intervals.

[0053] (2) Attitude Parameter Interface Unit: This unit is connected to the inertial measurement module and is used to input attitude parameters and complete the alignment and data supply with the event grouping time reference. This unit adopts a synchronous input method with time stamps and runs an attitude matching function internally: based on the attitude parameter time stamps and the grouped event set... The time coverage interval completes the parameter correspondence, ensuring that the processor processes each event set. The matching attitude parameters are read in real time; simultaneously, the spin angular velocity vector is extracted from the attitude parameters. It is then output to the processor to provide a real-time motion reference for rotation alignment compensation.

[0054] (3) Processor and Storage Unit: This unit is the core of online computing and includes a processor and memory. The processor receives packet event sets. and corresponding Run the rotation alignment compensation function afterward: for any event Calculate the time difference and make the rotation angle in for The length of the mold, Let be the unit rotation axis vector; construct the rotation correction matrix based on this.

[0055]

[0056] in It is the identity matrix. Unit rotation axis The antisymmetric matrix; using event camera calibration parameters to determine pixel coordinates Mapped to 3D points in the camera coordinate system Camera and press Three-dimensional points after compensation Then, the compensated 3D points Back projection onto the pixel plane yields the compensated pixel coordinates. The processor then runs the static / dynamic separation function within the compensation domain: [The function will then...] With the initial pixel coordinates Perform Euclidean distance comparison, if If it is a static event, it is considered a dynamic event; otherwise, it is considered a dynamic event. A preset pixel distance threshold is used to obtain the separated event set. The memory is used to cache grouped events, compensation results, and separated events, supporting the uninterrupted pipeline operation of the above processing chain.

[0057] (4) Subsequent localization and calculation processing of the processor and storage unit: After obtaining the separated event set, the processor continues to run the clustering and target matching localization function: the compensated pixel coordinates corresponding to the separated events are processed. Execution density clustering forms event clusters And calculate the centroid pixel coordinates.

[0058]

[0059] in The horizontal and vertical coordinates of the centroid pixel are used; the processor further extracts features such as the event cluster scale and performs nearest neighbor threshold matching with the target prior information injected before launch to determine the target event cluster and its image plane position.

[0060] (5) Target Information Output Unit: This unit is connected to the guidance and control system and is used to carry out target azimuth calculation and result transmission. The processor outputs the target information based on the target image plane position. The target's line-of-sight direction is calculated by combining the event camera calibration parameters, and the target's horizontal deflection angle is calculated relative to the current camera's optical axis. With pitch and yaw angle The target orientation information is obtained; the target information output unit outputs the target orientation information to the guidance and control system in real time with an update cycle consistent with the processor's cycle time, forming a closed-loop guidance.

[0061] The online link, composed of the aforementioned hardware units and their processing functions, enables the event set EEE to remain stable even under severe fluctuations in high-speed spin trigger rate, allowing for equal-beat grouping, synchronous data supply, real-time compensation separation, clustering and positioning, and output of target orientation information. This meets the engineering requirements for throughput, timing, and stability in high-speed spin guidance scenarios.

[0062] pass Figs. 1-3 The system structure and integrated online processing chain shown in this embodiment can stably access the event set under the conditions of high-speed spin and high dynamics of the carrier. It synchronously integrates attitude parameters and completes rotation alignment compensation, dynamic-static separation, clustering and target matching positioning, and target orientation calculation, thereby realizing the real-time output of target orientation information and meeting the engineering application requirements of guidance and control.

[0063] This implementation method unfolds step by step according to the working link of the seeker system, and describes the hardware configuration of the internal units of the event flow target detection module and the processing functions running on them. The flow of event data between units and the processing results are described mainly based on the real-time data path.

[0064] The accompanying drawings show a framework diagram and a flowchart of the seeker system and event stream target detection module of the present invention. It should be understood that each unit in the framework diagram and flowchart is implemented by hardware circuitry and its associated processing functions. These processing functions are executed by a processor to perform functions such as high-speed event stream access and caching, synchronous input of attitude parameters, rotation alignment compensation, static / dynamic separation, clustering and localization, and real-time output of target azimuth, thereby constituting the event stream target detection module and seeker system of the present invention.

Claims

1. An event camera seeker system for a high speed spinning carrier, characterized by, Comprise: Event camera sensor: for asynchronous perception of a scene under carrier spinning motion conditions, producing and outputting event stream data containing pixel coordinates, timestamps, and polarity, denoted as event set ; Inertial measurement module: fixedly installed inside the carrier, outputting attitude parameters, and synchronizing with event timestamps through time reference or alignment circuit shared with event camera sensor; Event stream target detection module: receives event set and pose parameters, the event set is processed according to preset timing Effective event screening, motion compensation, static and dynamic separation, target matching positioning and target azimuth information calculation operations are performed, target azimuth information is output, and target information is output to the guidance control system.

2. The nose system of claim 1, wherein Event stream target detection module arranged in the inner cabin of the seeker, at least comprising: Event stream receiving and buffering unit: for continuously accessing, buffering and time-stamp ordered outputting the event set ; Attitude parameter interface unit: for receiving attitude parameters and completing alignment with event time reference; Processor and storage unit: for calculating output target azimuth information according to event stream and aligned attitude parameters; Target information output unit, for outputting target azimuth information to the guidance control system.

3. The nose system of claim 2, wherein, The multi-level cache area of the event stream receiving and caching unit at least comprises on-chip cache and external high-speed storage cache, for avoiding packet loss and maintaining event output in monotonically increasing order when event stream rate fluctuates.

4. The nose system of claim 2, wherein, The time alignment circuit of the attitude parameter interface unit is used for mapping attitude parameter time markers to event time reference, so that each event packet reads the corresponding attitude parameters when processed by the processor.

5. A method of guiding using the guidance head system of any one of claims 1-4, characterized by, Comprise the following steps: Step (1): the event camera acquires raw data; Step (2): the processor reads event set E from the event stream receiving and caching unit, performs redundancy suppression and effective event screening grouping on the events; Step (3): the processor reads attitude parameters from the attitude parameter interface unit, performs rotation alignment compensation on each group of event samples to obtain compensated event pixel coordinates; separates static events from dynamic events to obtain separated event set; Step (4): clustering and target matching are performed on the separated event set to determine the corresponding target and output the matched target position; Step (5): according to the matched target position, the horizontal deflection angle and the pitch deflection angle of the target relative to the optical axis are calculated to obtain target azimuth information, which is output to the guidance control system in real time at a preset update period to form a closed loop.

6. The method of claim 5, wherein, The effective event screening grouping in step (2) is specifically: determining as an effective event when the event polarity belongs to the positive polarity set, the positive polarity set is , and grouping the effective event set at a fixed time interval to obtain a grouped event set ; wherein is a non-negative integer.

7. The method of claim 6, wherein, The rotation alignment compensation in step (3) is specifically: extracting a spin angular velocity vector from attitude parameters , the pixel coordinates of the event are denoted as , the time difference relative to the initial moment of the reference is calculated, and the rotation angle is denoted as ， wherein is the module length of is the unit rotation axis vector; Construct a rotation correction matrix: ， wherein is the identity matrix, is the identity rotation axis is the antisymmetric matrix of the identity rotation axis; Utilizing event camera parameters to map pixel coordinates to camera coordinate system three-dimensional points and obtain compensated three-dimensional points ;​ The compensated three-dimensional points are projected onto the pixel plane again to obtain compensated pixel point coordinates The compensated three-dimensional points are projected onto the pixel plane again to obtain compensated pixel point coordinates .

8. The method of claim 7, wherein, The separation of static events and dynamic events in step (3) is specifically: Within the compensation domain, the spatial consistency criterion is used for discrimination, and the initial time is determined. The event pixel coordinates are as And the compensated pixel coordinates and Calculate the Euclidean distance if it satisfies If it is a static event, it is considered a dynamic event; otherwise, it is considered a dynamic event. This is a preset pixel distance threshold.

9. The method of claim 8, wherein, Step (4) is specifically: compensated pixel point coordinates corresponding to the separated events forming event clusters by density clustering , calculating the event cluster centroid pixel point coordinates , satisfying: ， And the centroid pixel point coordinates and event cluster scale features are matched with the target prior information to determine the target event cluster and output the target image plane position.

10. The method of claim 1, wherein, Step (5) is specifically: Based on the target image plane position By combining the calibration parameters of the event camera, the target image plane position is converted into the target line of sight, and the horizontal deflection angle of the target relative to the optical axis is calculated from the line of sight. With pitch and yaw angle This serves as the output of target location information.

Images