基于太阳光谱特性的近红外机器视觉检测方法

Near-Infrared Machine Vision Detection Method Based on Solar Spectral Characteristics

摘要: 针对太阳光对机器视觉成像的干扰问题，本文提出一种基于太阳光谱特性的近红外视觉检测方法。该方法依据大气在近红外波段对太阳光的吸收特性，通过匹配特定波段的照明光源、窄带滤波片及工业相机，有效抑制太阳光对成像系统的干扰。具体而言，利用大气吸收作用降低工作波段内的太阳背景光，并借助光学滤波阻断其它波段的太阳辐射，从而提升成像质量与检测可靠性。实验表明，在太阳直射条件下，该方法能有效抑制光照不均与背景噪声，获得轮廓清晰的目标图像。曝光时间分析进一步证明，将曝光时间控制在1521 μs以内，太阳光干扰下成像的图像信噪比(Signal-to-Noise Ratio, SNR)可保持在20以上。本研究为复杂户外光照环境下实现稳定、低成本的机器视觉检测提供了一种有效的硬件解决方案。

Abstract: To address the interference of sunlight with machine vision imaging, this paper proposes a near-infrared machine vision detection method based on the spectral characteristics of sunlight. By leveraging the absorption properties of the atmosphere in the near-infrared band, the method can effectively suppress the interference of sunlight on the imaging system through the careful selection of illumination sources, narrow-band optical filters, and industrial cameras operating in specific wavelengths. Specifically, it utilizes atmospheric absorption to reduce solar background noise within the working waveband, while employing optical filtering to block solar radiation in other bands, thereby improving imaging quality and detection reliability. Experiments demonstrate that under direct sunlight, this approach effectively mitigates uneven illumination and background noise, producing clear images with well-defined contours. The exposure time analysis provides further evidence that an image Signal-to-Noise Ratio (SNR) greater than 20 under sunlight interference can be achieved by controlling the exposure time within 1521 μs. This study provides an effective hardware-based solution for achieving stable and low-cost machine vision detection in complex outdoor lighting environments.