Research On FPGA-Based Image Reconstruction Algorithm For Electrical Capacitance Tomography System

As a new technique under rapid development in recent years, electrical capacitance tomography technique (ECT) has become the mainstream development direction and research hotspot of process tomography technique. It has several advantages such as low-cost, non-invasive, good security performance, simple structure, wide application and etc, which has great development potential and broad industrial application prospect in addressing the issue of two-phase flow detection.As the key technology of ECT system research, image reconstruction algorithm has a significant impact on quality and speed of reconstructed image. Support vector machine(SVM) algorithm has more rigorous theoretical and mathematical foundations, no local minimum problems, good generalization ability and classification accuracy, so it is chosen to complete the image reconstruction. The main SVM calculations involve the"high-dimensional"calculation of input space, and this calculation will result in more memory-occupied and longer time-consumed. Generally, improved software algorithms can solve these problems, but, affected by serial-calculation nature of computer systems, this approach limits its applications in industrial occasions of real-time required. In this paper, SVM algorithm has implemented by FPGA, making full use of the parallel features of FPGA logic cells to achieve the purpose of improving running speed of algorithm.The hardware implementation of SVM based on FPGA is divided into two parts, sample-training and sample-decision. Among them, the training process uses gradient ascent algorithm which more suits to hardware implementation, and perfects the adjustment and determination of the threshold, so the optimal solution has a higher accuracy. Hardware structure of the whole process adopts the mode of serial calculation-parallel transmission, reducing the time-delay between each module, improving the operation speed. Decision process mainly produces the gray value of test samples. Meanwhile, it paves the way for image reconstruction. Taking 12-electrodes electrical capacitance tomography systems as research objects, it paper studied flow pattern identification of oil-water two-phase flow by establishing the mathematical model of the sensor working principle. SVM algorithm which based on the FPGA trains and classifies the standardization capacitance values, and then, carries on several simulation experiments about static property and the dynamic property. The experiment indicated that the proposed method was verified to have the advantages of minor error and high imaging quality, therefore provided a new idea for ECT image reconstruction.