Compressive sensing based image processing in trapview pest monitoring system

Missing information in image can be recovered by using the principles of lower sample rate methods, such as Compressive Sensing. This method can, at the same time, recover the missing information in the signal and do the compression of the original data. Lowering the sample rate is especially suitable for natural images in applications where minor visually loss of fidelity is acceptable. The goal is to achieve a substantial reduction in bit rate and image size. In this paper we analyze the performance and quality of Compressive Sensing approach applied on images captured by the TrapView automated camera station for pest detection. The reconstruction at the decoder side, if only small number of image samples is available, is tested in the paper. This is done with the goal to test different approach in image capturing - acquisition of only part of digital data, and then the reconstruction of the uncaptured/missing part, in order to obtain the original signal. This leads to decreasing the bit rate and transferred data volume through the mobile network, from the station to the TrapView cloud centre. It is shown that CS can provide a good quality image reconstruction with significantly reduced number of samples. The theory is tested using real images, obtained by the TrapView camera.