Imaging simulation of a dark field imperfection evaluation arrangement for spherical optical surface is proposed on the basis of reverse ray tracing. A typical evaluation arrangement for optical element surface imperfections seen in reflection under dark field condition is introduced. As for thin lens, it is a common case that light spots naturally emerge in the field of view (FOV) and dark field condition is no longer satisfied because multiple optical surfaces reflect rays into the camera lenses. A virtual arrangement model composed of a camera model, sample and light geometries, optical properties is established to further discuss the imaging of various lenses and imperfection geometries. The finite aperture camera model is adopted to generate abundant reverse rays for dark field scattering imaging of imperfections such as scratches and digs. But only a small portion of emitted rays whose paths can be traced to lights during propagation will contribute to the final image. Image simulation is performed by solving illuminance integral equation with Monte Carlo techniques. Simulated images of various lens geometries show the bright spots with respect to radius of curvature and centre thickness, which are numerically evaluated by our own defined spot coefficient. To eliminate the blind spot area caused by multiple reflections, lights are switched in turn and an image sequence is captured. A result image with reduced blind area is obtained by image fusion based on the accurate prediction of spot size and location for each frame in sequence. Imaging experiments of 4 thin lenses show that the distributions of spots correspond well with our simulation and proposed fusion method can reduce the blind area for spherical surface inspection while remains high sensitive for imperfections.

基于反向光线跟踪，提出了一种球形光学表面暗场缺陷评估装置的成像仿真。介绍了在暗场条件下反射中看到的光学元件表面缺陷的典型评估方案。对于薄镜头，通常的情况是，光点自然会在视场（FOV）中出现，并且不再满足暗视场条件，因为多个光学表面会将光线反射到相机镜头中。建立了由相机模型，样品和光的几何形状，光学特性组成的虚拟排列模型，以进一步讨论各种透镜和不完美的几何形状的成像。采用有限光圈相机模型来生成丰富的反向射线，以对瑕疵（例如划痕和凹陷）进行暗场散射成像。但是只有一小部分发出的光线在传播过程中可以追踪到光线，这些光线会有助于最终成像。通过使用蒙特卡洛技术求解照度积分方程来执行图像仿真。各种镜片几何形状的模拟图像显示了相对于曲率半径和中心厚度的亮点，这些亮点通过我们自己定义的光斑系数进行数值评估。为了消除由多次反射引起的盲区，请依次切换光线并捕获图像序列。基于对每个帧顺序的光点大小和位置的准确预测，通过图像融合获得具有减少的盲区的结果图像。