摘要
目前,ToF(Time of Flight)三维成像技术在人脸检测、3D目标识别、三维重建等视觉任务领域具有广阔的应用前景。然而,用ToF相机所获得的深度信息往往存在与像素、温度、深度畸变、多径干扰以及背景光相关的噪声干扰。现有的ToF优化算法耗时较大且很难保留目标的细节信息,这些问题严重影响了ToF相机的实际应用。针对以上问题,本文提出一种实时的基于振幅图的ToF深度图优化方法。首先通过ToF接收端采集的原始数据生成带有噪声的振幅图像。针对振幅图中的噪声,选用快速高效的双边网格滤波对振幅图进行去噪。然后,利用优化后的振幅图生成掩码以分割出深度图中前景和背景区域。同时,对深度图中的噪声以及误差像素用滤波的方式优化,最后将优化后的深度图和掩码融合生成最终的深度图。实验结果表明,本文所提算法可以实时有效地滤除深度图噪声,去除背景噪声的干扰,同时能很好地保留深度图中目标对象的细节信息。有助于ToF相机拥有更广泛的应用场景。
At present,ToF(Time of Flight)3D imaging technology has broad application prospects in visual tasks such as face detection,3D object recognition,and 3D reconstruction.However,the depth information obtained with ToF cameras often suffers from noise interference related to pixels,temperature,depth distortion,multipath interference,and background light.The existing ToF optimization algorithms are time-consuming and difficult to preserve the detailed information of the target,which seriously affects the practical application of ToF cameras.In response to the above issues,this article proposes a real-time ToF depth map optimization method based on amplitude maps.Firstly,amplitude images with noise are generated from the raw data collected by the ToF receiver.For the noise in the amplitude map,a fast and efficient bilateral grid filter is selected to denoise the amplitude map.Then,the optimized amplitude map is used to generate a mask to segment the foreground and background regions in the depth map.At the same time,the noise and error pixels in the depth map are optimized by filtering,and finally,the optimized depth map and mask are fused to generate the final depth map.The experimental results show that the algorithm proposed in this paper can effectively filter out depth map noise in realtime,remove background noise interference,and preserve the detailed information of the target object in the depth map.Helping ToF cameras have a wider range of application scenarios.
作者
葛晨阳
姜少敏
刘洋
邓鹏超
GE Chenyang;JIANG Shaomin;LIU Yang;DENG Pengchao(Institute of Artificial Intelligence and Robotics,Xi'an Jiaotong University,Xi'an 710049,China)
出处
《微电子学与计算机》
2024年第6期57-64,共8页
Microelectronics & Computer
基金
国家自然科学基金(61627811)
陕西省自然科学基金(2021JZ-04)
陕西省重点研发高校联合项目(2021GXLH-Z-093)
陕西省技术创新引导专项(2021QFY01-03)。