Visual Ranging Based on Object Detection Bounding Box Optimization

Faster and more accurate ranging can be achieved by combining the object detection technique based on deep learning with conventional visual ranging. However, changes in scene, uneven lighting, fuzzy boundaries other factors may result a non-fit phenomenon between bounding box object. The pixel spacing cause errors. To reduce spacing, increase degree of fit object, improve accuracy, an optimization method is proposed. Two evaluation indicators, WOV HOV, are also proposed to evaluate results optimization. experimental show that width optimized 1.19~19.24% height 0~12.14%. At same time, experiments demonstrate improves accuracy. In addition, few practical monocular range measurement techniques determine distance whose size unknown. Therefore, similar triangle difference suggested measure items unknown size. A experiment carried out detecting box, reveal relative error within 6 m 0.7% 2.47%, allowing for precise measurement.