基于代数多重网格的图像传感器物体识别技术

摘要

物联网中的物体识别可以减少人为的参与，提高物物相连的效率。该文针对物联网环境中的物体识别进行了初步研究，提出了一种结合代数多重网格的物体识别的方法，降低了物理存储和网络传输的代价。首先运用代数多重网格方法对不同模糊程度的图像进行重建，在此基础上进行特征检测；然后运用“词袋”模型对使用了代数多重网格方法与未使用该方法的物体识别进行了对比试验。实验结果表明，运用一定程度的模糊图像识别物体能得到较高的稳定性，并且提升了与非同一场景的物体识别的区分度；运用代数多重网格方法的“词袋”模型提高了物体识别的准确率。%Object recognition in the Internet of things (IOT) can make the connection of objects easier by reducing the participation of the people significantly. Because of the particularity of IOT, how to reduce the storage and network transmission cost is an important research topic. In this paper, algebraic multigrid method is proposed to reduce the storage and network transmission costs in the application of object recognition under the environment of IOT. Firstly, the coarse grid data extracted by algebraic multi-grid (AMG) method is reconstructed, then the features are detected for object recognition, and finally, an object recognition experiment is provided by the "bag of words" model in the images reconstructed with and without the algebraic multi-grid method. The experimental results show that the "bag of words" model with algebraic multi-grid method can recognize the blurred objects more steadily than the model without algebraic multi-grid method, and the distinguish degree is improved between the same scenes and the different ones by the method of AMG. Therefore, AMG method can be used as a new feature extraction method in object recognition under the IOT environment.