dc.description.abstract | In recent years, most of the machine vision systems use embedded hardware and software co-design, which uses FPGA to replace some DSP modules. In traditional development, we first design software prototype and then choose some modules that designed by hardware. We use a series designed to connect this hardware modules to compose hardware architecture. However, this type of architecture lacks flexibility. If system requirements just need to do a little change, the architecture need to whole redesign.
In this thesis, we propose embedded vision hardware and software co-design methodology. Firstly, we analyze the system requirements with IDEF0. This way is analysis whole system hierarchically and divided into many modules. Secondly, we use Grafcet establish discrete event model for every modules. Then we through the way of Grafcet synthesis to produce software code and hardware design. This development approach needs not coding in prophase of system design. Additionally, we design a hardware interface controller, which is suitable in hardware and software co-design architecture. This controller contains all of hardware modules, and designers can select desired target modules according to system requirements which include efficacy, elasticity, cost, and power consumption. It is not necessary to redesign hardware architecture, the designer just to change the order of hardware modules through software. The hardware interface controller can increased development flexibility, and accelerate the system development process.
| en_US |