Scholarlink

<p>Event-based Vision Sensors (EVS) have been increasingly utilized for Space Domain Awareness (SDA) applications over the past several years. Newer models demonstrate improved performance compared to early generation systems; however, the absolute limits of their sensitivity with respect to fundamental physical parameters (i.e. on-chip photon flux) for detecting dim point source objects has not been fully quantified. Additionally, the complex exercise of optimizing tunable sensor biases for SDA has not been explored for the newest generation sensors. This work shows that with optimized biasing, scan rate, and a novel denoising strategy, the IMX636 sensor is capable of detecting point sources that produce on the order of several hundred photons per second on the focal plane. Our detailed characterization allows users of the technology to predict detection limits when paired with varied optical systems. The result is included in an open-source software tool. Additionally, bias optimization is shown to extend sensitivity by 1.05 visual magnitudes, <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\({{m}}_v\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>m</mi> <mi>v</mi> </msub> </math></EquationSource> </InlineEquation>, and when coupled with the our denoising method, enables detection of <InlineEquation ID="IEq5"> <EquationSource Format="TEX">\(2.9\times \)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>2.9</mn> <mo>×</mo> </mrow> </math></EquationSource> </InlineEquation> more star streaks compared with default biases. The detailed modeling of pixel level performance further enables a method for reconstructing object brightness from event camera output by calibrating the sensor’s dark current and incorporating a model that accounts for the pixel’s illumination-dependent temporal response and logarithmic compression. Ultimately, the ability to accurately extract photometric information from an event camera may eventually extend to other characterization methods including polarimetry and spectral analysis with proven EVS benefits of excellent temporal resolution, low power, sparse data output.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Advanced EVS Pixel Modeling to Improve Dim RSO Detection up to 2.9\(\times \) and Enable Accurate Photometric Reconstruction for SDA

  • 期刊论文

  • 发表日期：

    2026年7月28日

查看全文
  • Brian McReynolds,
  • Rachel Oliver,
  • Zachry Theis,
  • Peter McMahon-Crabtree,
  • Michal Zolnowski,
  • Tobi Delbruck

摘要

Event-based Vision Sensors (EVS) have been increasingly utilized for Space Domain Awareness (SDA) applications over the past several years. Newer models demonstrate improved performance compared to early generation systems; however, the absolute limits of their sensitivity with respect to fundamental physical parameters (i.e. on-chip photon flux) for detecting dim point source objects has not been fully quantified. Additionally, the complex exercise of optimizing tunable sensor biases for SDA has not been explored for the newest generation sensors. This work shows that with optimized biasing, scan rate, and a novel denoising strategy, the IMX636 sensor is capable of detecting point sources that produce on the order of several hundred photons per second on the focal plane. Our detailed characterization allows users of the technology to predict detection limits when paired with varied optical systems. The result is included in an open-source software tool. Additionally, bias optimization is shown to extend sensitivity by 1.05 visual magnitudes, \({{m}}_v\) m v , and when coupled with the our denoising method, enables detection of \(2.9\times \) 2.9 × more star streaks compared with default biases. The detailed modeling of pixel level performance further enables a method for reconstructing object brightness from event camera output by calibrating the sensor’s dark current and incorporating a model that accounts for the pixel’s illumination-dependent temporal response and logarithmic compression. Ultimately, the ability to accurately extract photometric information from an event camera may eventually extend to other characterization methods including polarimetry and spectral analysis with proven EVS benefits of excellent temporal resolution, low power, sparse data output.