“多源异构感知对象快速精准的分割、检测 、定位、跟踪和识别”关键考核指标第三方检测数据

面向无人系统地外探测任务对复杂无约束环境精准感知与透彻理解的需求，课题二针对地外星表形貌原始自然、纹理相似度高、目标先验知识欠缺等特点，开展复杂场景下感知对象的精准分割、快速检测与识别、目标定位与快速连续跟踪等方法研究，突破相似非结构化地形的精准分割、复杂场景先验知识欠缺条件下对象特征挖掘与精准辨识等关键技术，构建地外场景标注样本集，形成多维特征融合的学习训练平台，实现复杂场景下多源异构感知对象的精准辨识。 针对项目及课题验收对多源异构感知对象快速精准的分割、检测、定位、跟踪和识别的具体要求，本文件明确了测试目的，给出了指标要求，并详细介绍了测试的硬件环境，对各项功能和指标测试给出了具体要求。To address the demands of precise perception and thorough understanding of complex unconstrained environments for extraterrestrial exploration missions using unmanned systems, Task 2 conducts research on methods including precise segmentation, rapid detection and recognition of perception objects, as well as target localization and rapid continuous tracking in complex scenarios, targeting the characteristics of extraterrestrial planetary surfaces such as original and natural morphology, high texture similarity, and lack of prior knowledge about targets. It breaks through key technologies such as precise segmentation of similar unstructured terrain, object feature mining and precise recognition under the condition of insufficient prior knowledge in complex scenarios, constructs an annotated sample dataset for extraterrestrial scenarios, develops a learning and training platform integrating multi-dimensional features, and achieves precise recognition of multi-source heterogeneous perception objects in complex scenarios. In response to the specific requirements for rapid and precise segmentation, detection, localization, tracking and recognition of multi-source heterogeneous perception objects for project and task acceptance, this document clarifies the test objectives, specifies indicator requirements, elaborates on the hardware environment for testing, and specifies detailed requirements for each functional and indicator test.

搜集汇总

背景与挑战

背景概述

该数据集是为‘地外探测无人系统自主智能精准感知与操控’项目的验收而提供的第三方检测数据，针对多源异构感知对象在复杂场景下的快速精准分割、检测、定位、跟踪和识别等关键考核指标。它支持地外探测任务中无人系统对相似非结构化地形和先验知识欠缺环境的感知能力评估。