Scholarlink

<p>A novel obstacle detection method for unmanned surface vehicles (USVs) based on stereo vision and 3D lidar Bayesian fusion is proposed for reliable close-range detection. The USV obstacle environment is modeled as a 2D grid map, with decision information from stereo vision and 3D lidar used to determine grid attributes. A method to determine the conditional probability distribution of each decision information under the grid attributes is proposed based on sensor measurement range and accuracy. It is assumed that the prior probabilities for obstacle and passable areas are both 0.5. Using Bayesian probability formula, posterior probabilities are calculated, and grid attributes are determined according to the maximum posterior probability criterion. The fusion grid representation result for USV obstacles is then obtained. In the method verification experiment, sensor data from four representative sea obstacle scenes were used. Comparison results show that the proposed fusion detection method effectively leverages the complementary characteristics of the sensors, significantly reducing false alarms from individual sensors.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel obstacle fusion detection approach for unmanned surface vehicle based on stereo vision and 3D lidar

  • 期刊论文

  • 发表日期：

    2025年5月12日

查看全文
  • Deqing Liu,
  • Jie Zhang,
  • Jiucai Jin,
  • Yi Ma,
  • Bing Zheng,
  • Ligang Li

摘要

A novel obstacle detection method for unmanned surface vehicles (USVs) based on stereo vision and 3D lidar Bayesian fusion is proposed for reliable close-range detection. The USV obstacle environment is modeled as a 2D grid map, with decision information from stereo vision and 3D lidar used to determine grid attributes. A method to determine the conditional probability distribution of each decision information under the grid attributes is proposed based on sensor measurement range and accuracy. It is assumed that the prior probabilities for obstacle and passable areas are both 0.5. Using Bayesian probability formula, posterior probabilities are calculated, and grid attributes are determined according to the maximum posterior probability criterion. The fusion grid representation result for USV obstacles is then obtained. In the method verification experiment, sensor data from four representative sea obstacle scenes were used. Comparison results show that the proposed fusion detection method effectively leverages the complementary characteristics of the sensors, significantly reducing false alarms from individual sensors.