基于压缩传感的交通流量数据压缩方法

为准确获得用于数据压缩的变换矩阵,引入了基于压缩传感的交通流量数据压缩方法,在数据压缩端无需考虑变换矩阵的选择问题,直接通过高斯投影实现高效数据压缩。首先验证了交通流量数据在经过K-SVD方法训练过的字典上能够实现稀疏表达;然后在数据压缩端,通过具有限制性等距条件的随机矩阵将原始高维数据投影到低维空间上,实现数据的高效快速压缩;最后在数据传输后,通过凸优化算法在交通信息处理端完成数据解压缩。以美国某高速公路线圈传感器采集到的交通流量数据,对本文方法进行了验证。试验证明:该方法能够实现快速高效的压缩编码,当压缩比为4∶1时,解压缩相对误差仅为0.060 8。

Compression method of traffic flow data based on compressed sensing

In order to obtain transformation matrix accurately,a new compression method of traffic flow data based on compressed sensing was introduced.The original data were projected into the low-dimension space directly by Gauss projection regardless of transformation matrix selection at the data compression side.Firstly,traffic flow data were proved to have sparse representation under the K-SVD trained dictionary.Secondly,original high-dimension data were projected into low-dimension space at the data compression side by using the random matrix with restricted isometry property,which made efficient and rapid data compression possible.Finally,after data transmission,data decompression were accomplished by convex algorithm at the data processing side.The traffic flow data obtained from the coil sensors located on a certain highway of America were used to validated the new method.The experimental result shows that the data compression method is fast and efficient.When the compression ratio is 4∶1,the relative error of data decompression is only 0.060 8.5 tabs,8 figs,18 refs.