将非相似成像理论引入到立体视觉理论中, 探索了一种非相似立体视觉机制, 建立了空间目标三维定位模型和深度精度模型。根据流行的非相似等距投影理论, 对非相似立体视觉三维定位方程进行了数理推导, 并实验验证了所建模型的有效性。结果表明: 利用三维定位模型可以准确地获取大空域场景中不同深度的多个目标位置信息, 定位误差随着目标深度的增大而缓慢变大, 在30 m深度处的定位误差达到1.32 m, 误差变化规律与深度精度模型结果吻合良好, 且误差值远小于相似立体视觉模型的定位误差。所建模型的优势是无需对非相似畸变图像进行校正, 便可较好地得到空间目标的三维坐标, 研究工作对于拓展非相似成像理论在大空域态势感知、目标侦察等领域的应用具有重要意义。

By introducing the popular equidistant projection theory, a non-similar stereo vision was explored, and the three-dimensional (3D) location model and depth precision model were established. The correctness of proposed models was verified through carring out the 3D location experiment with a 115°×90° infrared stereo vision system. From the experimental results, the targets in different depths and large-airspace scene were located successfully. The depth error increases smoothly with increasing target depth, and it reached 1.32 m in the depth of 30 m. The error variation of 3D location model matched well with the results of depth precision model, and the whole error was obviously less than that of similar stereo vision model. The advantages of the proposed model is that the 3D coordinates of space target can be obtained directly without correcting the non-similar distorted images, and it contributes to enriching and improving the theory of stereo vision, and will help to spread the application of non-similar imaging theory in the fields of large-airspace situation awareness, target detection and so on.