An improved vision-based wastewater velocity measurement system using discontinuity-preserving smoothing and GPU acceleration

Automatic long-term measuring wastewater velocity is an important and challenging task in hydraulic systems. This paper proposed a vision-based wastewater velocity measurement method using Bilateral filter that is a discontinuity-preserving smoothing as a prior-processing step. Experimental results showed that using Bilateral filter can improve estimation accuracy over existing methods. An effective background creation algorithm and simple floating waste tracking algorithm based on binary blob properties are also discussed in this paper. Furthermore, by implementing the proposed method on massively parallel GPU (graphics processing units) using the CUDA (compute unified device architecture) programming model, we can achieve a satisfactory acceleration to apply in real-time applications. Memory usage optimization methods are discussed and analyzed for effective implementation in graphics hardware.