太赫兹科学与电子信息学报, 2017, 15 (3): 459, 网络出版: 2017-11-27
基于McFW的图像处理算法验证框架设计
Design of image processing algorithm testing framework based on McFW
McFW框架 视频监控 算法验证 图像处理 嵌入式开发 DM8168处理器 Multi-channel FrameWork video surveillance algorithm testing image processing embedded development DM8168
摘要
借助DM8168硬件平台, 在 McFW框架下, 使用 Link API接口, 研究设计了具备灵活性和实用性的图像处理算法验证框架。该框架逻辑清晰, 结构紧凑, 充分利用DM8168处理器的多核和多通道特性, 克服了原生 McFW框架中 DSP图像处理算法只应用于摄像头捕获数据的局限性, 着重于满足图像处理算法验证, 开发各项需求, 方便算法开发人员移植、集成和测试图像处理算法。该框架既适用于算法验证, 又可用于嵌入式产品开发, 为图像处理算法由PC机向嵌入式移植提供了很好的过渡。最后基于该框架初步实现了均值偏移目标跟踪算法的移植, 在多路摄像头输入的情况下, 针对固定场景视频监控, 满足实时性要求, 硬件资源分配合理, 证明了该算法验证框架的正确性和有效性。
Abstract
An image processing algorithm testing framework is designed, which is extensible and practical by using Link APIs of Multi-channel Frame Work(McFW) based on DM8168 System On Chip(SOC). This testing framework is logical and structured, and also makes full use of the multi-core and multi-channel in DM8168 SOC, which overcomes the limitation of mere camera data source feeding to the DSP core in the primitive McFW APIs. Since it focuses on meeting various requirements of imageprocessing algorithm testing and development, it truly contributes to transplanting, integrating or testing image processing algorithms. The framework can be served as a solution for embedded product development as well as for algorithm testing. It also helps to make the transition from Personal Computer(PC) development to embedded platform development smoothly. The object tracking algorithm is transplanted based on means shift theory, which works well in settled multi-channel video surveillance scenarios. Embedded in the framework designed in this work, the image processing meets real-time requirements with sensible resource allocation, which proves its effectiveness.
沈宝成, 廖庆敏, 王德生. 基于McFW的图像处理算法验证框架设计[J]. 太赫兹科学与电子信息学报, 2017, 15(3): 459. SHEN Baocheng, LIAO Qingmin, WANG Desheng. Design of image processing algorithm testing framework based on McFW[J]. Journal of terahertz science and electronic information technology, 2017, 15(3): 459.