Outdoor long-range target positioning based on multi-monocular vision fusion

This paper proposes a target positioning algorithm based on multi-monocular vision fusion. Monoculars with different focal lengths are used to form a measurement system, and the images obtained by monoculars are fused and deduplicated.Because of the fusion of multi-monocular detection results, it can accurately detect targets at different distances, thereby expanding the single-lens field of view and reducing the impact of target detection error. At the same time, the rapid target detection feature of yolov3 can be used to measure multiple targets, which can be applied to a variety of outdoor positioning scenarios, such as pedestrian measurement and vehicle distance measurement in unmanned driving. In the outdoor experiment part, we introduced the outdoor multi-monocular over-the-horizon positioning device, and measured the distance of the car as far as 135 meters, with an average error of 5.4%.