CN-122002146-A - Event data set construction method and system

Abstract

The invention relates to the technical field of data processing, in particular to an event data set construction method and system, which are used for acquiring event data of a target area from output data of one event camera and acquiring event data of the target area of other event cameras in a target scene by utilizing pixel size multiples among a plurality of event cameras. The invention constructs the data set which can truly reflect the scene and contains event data with different resolutions by utilizing the optical light splitting method, and improves the accuracy and the simplicity of the acquisition of the data set.

Inventors

• ZHANG YISA
• Lv Hengyi
• ZHAO YUCHEN
• SUN MING
• FENG YANG

Assignees

• 中国科学院长春光学精密机械与物理研究所

Dates

Publication Date

20260508

Application Date

20260409

Claims (6)

1. 1. A method of event data set construction, comprising: S1, utilizing a half-reflecting half-lens to project optical information of a target scene into a plurality of event cameras with different pixel sizes; S2, establishing a coordinate space mapping relation between one event camera and other event cameras according to the multiple of the pixel sizes among the event cameras; And S3, determining imaging areas corresponding to the imaging areas of one event camera in other event cameras by utilizing a coordinate space mapping relation, and collecting event data in the imaging areas of all event cameras to obtain event data sets with different resolutions.
2. 2. The event data set construction method according to claim 1, wherein each event camera in step S1 is equipped with an adjustable-focus optical lens.
3. 3. The method of claim 2, wherein if the pixel sizes of the plurality of event cameras are the same, adjusting the focal length of the lens of each event camera so that the resolutions of the images acquired by the plurality of event cameras are different.
4. 4. An event dataset construction system, comprising: A plurality of event cameras of different pixel sizes; The half-reflecting half-lens is used for projecting optical information of a target scene into a plurality of event cameras; The control center is used for controlling the event cameras to shoot a target scene at the same time, establishing a coordinate space mapping relation between one event camera and other event cameras according to the multiple of pixel sizes among the event cameras, determining imaging areas corresponding to the imaging areas of one event camera in the other event cameras by utilizing the coordinate space mapping relation, and collecting event data in the imaging areas of all event cameras to obtain event data sets with different resolutions.
5. 5. The event dataset construction system as claimed in claim 4, wherein an adjustable-focus optical lens is provided at each event camera.
6. 6. The event dataset construction system as claimed in claim 4, wherein the control center is connected by a cable and controls each event camera.

Description

Event data set construction method and system Technical Field The invention belongs to the technical field of data processing, and particularly relates to an event data set construction method and system. Background The super-resolution of the image is to reconstruct a high-resolution image from a low-resolution image to recover the details, textures and edge information lost in the low-resolution image due to insufficient sampling, and the super-resolution of the event data is performed not only in the spatial domain but also in the time domain. The event camera is an image acquisition device which takes a dynamic vision sensor (also called an event trigger detector, a nerve morphology detector and a bionic vision sensor) as a core imaging device to acquire light intensity change information in a scene. The event camera has been widely used in the machine vision fields such as object classification and recognition by virtue of the advantages of large dynamic range, high time resolution, less redundant data and the like. In order to improve the recognition accuracy of the event camera in the field of target classification recognition, a part of scientific researchers perform super-resolution research on the output data of the event camera and construct a data set. However, because of the difficulty in making the data set, the current method for applying and verifying the super-resolution algorithm of the event camera has fewer data sets, namely the low-resolution event data is obtained by downsampling the high-resolution event data, the data set obtained by the method cannot truly reflect the sampling results of the event camera on scenes under different resolutions, and the downsampling method often has a certain regularity and cannot ensure that a downsampling model is reversely applied to the super-resolution algorithm. Disclosure of Invention Therefore, the invention aims to provide a method and a system for constructing an event data set, and provides a manufacturing method capable of truly reflecting the sampling result of an event camera for constructing the data set of the super-resolution algorithm of the event camera by utilizing an optical spectroscopic method. In order to achieve the above purpose, the technical scheme of the invention is realized as follows: a method of event dataset construction, comprising: S1, utilizing a half-reflecting half-lens to project optical information of a target scene into a plurality of event cameras with different pixel sizes; S2, establishing a coordinate space mapping relation between one event camera and other event cameras according to the multiple of the pixel sizes among the event cameras; And S3, determining imaging areas corresponding to the imaging areas of one event camera in other event cameras by utilizing a coordinate space mapping relation, and collecting event data in the imaging areas of all event cameras to obtain event data sets with different resolutions. Further, each event camera in step S1 is equipped with an adjustable-focus optical lens. Further, if the pixel sizes of the event cameras are the same, the lens focal length of each event camera is adjusted so that the resolutions of the images acquired by the event cameras are different. The event data set construction system comprises a plurality of event cameras with different pixel sizes, a half-reflection half-lens used for projecting optical information of a target scene into the plurality of event cameras, a control center used for controlling the plurality of event cameras to shoot the target scene at the same time, establishing a coordinate space mapping relation between one event camera and other event cameras according to pixel size multiples among the plurality of event cameras, determining imaging areas corresponding to imaging areas of one event camera in the other event cameras by utilizing the coordinate space mapping relation, and collecting event data in the imaging areas of all event cameras to obtain event data sets with different resolutions. Further, an adjustable-focus optical lens is equipped at each event camera. Further, the control center is connected through a cable and controls each event camera. Compared with the prior art, the invention has the following beneficial effects: The invention creates the method and the system for constructing the event data set, and utilizes a half-reflection half-mirror to construct an optical common-path structure, so that a plurality of event cameras with different pixel sizes receive optical information of the same physical scene, on the premise of ensuring spatial alignment on a physical layer, the pixel size ratio is used as a spatial scale mapping coefficient on the basis, and a deterministic mapping relation among a plurality of event camera coordinate systems is established. The need for post-processing registration is fundamentally avoided. The invention does not adopt simple central area cutting, but accurately maps the se