There are many embedded vision board out there. You have the choice of using general purpose processors (e.g Intel Core i5, Atom), DSP processors (e.g. TI Davinci series), custom implementation with IPs (e.g FPGA) or hybrid systems (FPGA+ ARM, x86+GPU etc.)
Before choosing a suitable platform, first you need to profile your algorithm and determine your time constraint if any. Do you need real-time processing? Next is to determine the max power consumption acceptable in your embedded system. Is your embedded system tethered to power supply? Running on batteries? Require long battery life etc.?
From there on, you can narrow down to what platform is suitable. It all comes down to the complexity of your algorithm, the need for real-time or not, and the power consumption desirable in the end product.
You should also consider the cost (GPP are cheaper), time to develop (FPGA takes longer time to develop), as well interface required (USB vision, GigE, CameraLink etc).
In the market, there are several embedded vision solutions from National Instruments (smart cameras), Adlink (e.g. Neon series, embedded vision board) to name a few.
To achieve fast results that are essentially looked for, there are many customized 2D digital signal processors. these DSPs integrate real time routines.
Take a look on the attached file. but there are many other technologies;
There are many embedded vision board out there. You have the choice of using general purpose processors (e.g Intel Core i5, Atom), DSP processors (e.g. TI Davinci series), custom implementation with IPs (e.g FPGA) or hybrid systems (FPGA+ ARM, x86+GPU etc.)
Before choosing a suitable platform, first you need to profile your algorithm and determine your time constraint if any. Do you need real-time processing? Next is to determine the max power consumption acceptable in your embedded system. Is your embedded system tethered to power supply? Running on batteries? Require long battery life etc.?
From there on, you can narrow down to what platform is suitable. It all comes down to the complexity of your algorithm, the need for real-time or not, and the power consumption desirable in the end product.
You should also consider the cost (GPP are cheaper), time to develop (FPGA takes longer time to develop), as well interface required (USB vision, GigE, CameraLink etc).
In the market, there are several embedded vision solutions from National Instruments (smart cameras), Adlink (e.g. Neon series, embedded vision board) to name a few.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing