项目名称: 基于二维和三维数据融合的室内物体识别方法研究
项目编号: No.61202158
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 于仕琪
作者单位: 深圳大学
项目金额: 24万元
中文摘要: 物体识别是人工智能领域中的一项核心技术,是机器人感知环境的必备功能。目前无论基于二维图像,还是基于三维点云的物体识别技术,都面临一些难以克服的困难。最近出现了一种新型传感器(微软Kinect),它可以同步采集二维彩色图像和三维深度图像。利用这种新型传感器来识别物体具有其他传感器所没有的优势。 本项目将研究二维图像与三维深度数据融合的特征提取方法,借鉴人脑的认知规律,设计出具有区分能力的特征;同时针对融合的数据,设计机器学习方法,使之能够根据二维和三维信息,自动学习出物体的姿态,以及物体之间的关系,提升物体识别的性能,提升计算机对环境的理解能力。为物体识别这一经典问题提出新的解决方案,这将为家庭服务机器人、无人驾驶汽车等应用提供理论指导和技术基础。
中文关键词: 物体分类;物体检测;深度图;;
英文摘要: Object recognition in computer vision is the task of finding a given object in an image or video sequence. It is the key component in an artificial intelligence system. There are some problems which are hard to solve in object recognition whatever using 2D images or 3D point clouds. A new kind of sensor (Kinect by Microsoft Corp.) has been developed in the recent two years. It can capture 2D color images and 3D depth images, and it is becoming a popular sensor for object recognition. We will develop object recognition methods based on the fusion of 2D color images and 3D depth image. With the help of knowledge from cognitive science, we will design features with great discriminant. The advance learning algorithms are also what we concern because it is important in handling pose variation and object position relations. Our final aim is to improve the perception of the computer. The algorithms developed by us will help household service robots, driverless cars improve their intelligence.
英文关键词: object classification;object detection;depth images;;