Target Recognition in Turbulent Diffusion Channels

Target recognition in fluidic environment is an important research topic which can be implemented in various applications. Turbulent diffusion is a common phenomenon in this scenario, while the turbulent diffusion channels are not well-understood. In this letter, a BP neural network based on Bayesian regularization (BR-BPNN) is proposed to recognize a target in a bounded turbulent diffusion environment. Besides, we first propose the concentration total variation comparison (CTVC) algorithm to select the dominant sensing points in the channel. The results show that the radius and location of the target are recognized precisely, and the recognition accuracy is guaranteed (e.g., using 20 dominant sensing points, the error rate of radius is 1.73%, and the error rate of location is 0.97%).