为使靶场光测设备在进行基于图像的自动调焦过程中建立适当的调焦窗口以降低计算量并克服背景干扰，提出了结合视觉感知机制的调焦窗口构建方法。对图像进行金字塔分层获得多层子图；采用Gabor滤波器和基于高斯差分(DoG)的模型模拟人类大脑视皮层细胞功能，对子图层进行信息提取和特征融合，获得基于视觉感知的特征图；对该特征图进行阈值去噪，以减少干扰，并以边界扩展法建立适合于后续处理的矩形调焦窗口。实验结果证明，对于各种调焦状态下的图像，尤其是深度离焦状态，都能快速、准确地针对目标构建适当的调焦窗口，处理720 pixel×576 pixel的目标图像耗时小于130 ms。该方法能够满足靶场的实际要求，为整个调焦过程实时建立稳定、准确的调焦窗口，具有广泛的工程应用价值。

In order to build right focus window for optical measure equipment in shooting range, a novel method combined with visual perception is proposed. It can cut down compute time and overcome disturbance of background during auto-focus procedure based on image. The image is separated into several sub-images by using pyramid structure. The function and principle of cells in human visual cortex are simulated by Gabor filter and difference of Gaussian (DoG) model. Feature maps from the sub-images are extracted and combined together to get final feature map based on visual perception. The threshold is used to denoise disturbance in the feature map and build regular rectangle focus window area by spreading boundary. After plenty of experiments，the results show that this method can build focus window rapidly with exact size and right position for the target in image. And it is available for every kinds of focus states, especially for deep-defocus state. To deal with the image of 720 pixel×576 pixel, the compute time is less than 130 ms. The proposed method can construct steady and exact focus window during the whole focus procedure, which satisfies the practical demand of shooting range and is of great and wide value for engineering application.