Article Preview
TopIntroduction
The IoT (Internet of Things) has become the massive driven force in the area of sensor computing platforms especially in sectors like commercial and industrial domains (Derhamy, Eliasson, Delsing, & Priller, 2015). It has a wide range in different applications like cameras, home automation, drones, health monitors, smart grid, logistics, transportation, agriculture, manufacturing, and many more. IoT-based devices can also perform local computation and have intelligence integrated with edge devices which allow us to make a global decision depends on local sensor measurements (Ray, 2016). These are applications that boost the intelligence of operations. The study of the market and the analysis of IoT applications have created substantial growth in both areas that are sales and cloud computing services. Even these IoT devices have established the largest market for semiconductor electronics and the software can range from 8-bit microprocessor applications to highly optimized systems which can include several cloud-based integrated microprocessors. These types of applications can have complex configurations also as battery power and operation constrained, performance saturation level, issues related size and weight of the devices, location, and installation constraints, operation on wired power, cost-constrained that might be related to the addition of different applications and many more. Still, many standardized platforms are generating low power and performance but the need for high-performance platforms is in very high demand because nowadays there is a large amount of data and this data needs high computation that can generate intelligence into the edge devices. Normally the platforms used for this IoT implementation are Graphics Processing Units (GPUs)], Field-Programmable gate array (FPGA), or some application-based integrated circuit implementation (Polianytsia, Starkova, & Herasymenko, 2017). These platforms can improve the performance depend on the application but at the same time, IoT has to balance power, size, performance as well as cost also. In recent times there is a vast change in FPGA programming which makes it more feasible and efficient for IoT implementation.
Nowadays most of the IoT devices are using CPU based platforms but they are also gradually moving towards GPU, FPGA, and other embedded platforms. These platforms are application-oriented because whenever there is a need for improved efficiency, custom computation, power, size, etc. developers need a more aggressive platform with custom features. There is a lot of design issues related to platform design mainly in terms of its architecture and cost. Automation has played an important role during the designing phase which makes it more feasible on the basis of lower risk and cost. Despite all these advantages in working with custom platforms, there are different types of challenging and complex design flows that have become a barrier during the adoption of custom platforms especially in association with hardware accelerators. The design of space exploration and the achievement of goals that help to reduce the design costs, the implementation time, and risk that could ensure both technological and commercial success also played an important role which can be done by computer-aided design.
This paper will represent the generic IoT device design process and the different types of platforms available for IoT devices that could minimize the cost, power, volume, and enhance the performance efficiently. Representation of the design process will be done through application based on computer vision. Computer vision algorithms are widely used today and these are considered to be the key aspects for intelligence in edge devices. Vision applications are more data-intensive, high communication delay, and transforming the video data onto the centralized cloud computing have some limitations. These are some of the reasons due to which vision system is rapidly moving into IoT applications and analysts have predicted that connected imaging equipment will increase by 17% annually in the year 2020.
A lot of local computing is needed for IoT applications in computer vision to achieve high-scale performance and scalability in monitoring systems, body cameras, public transport systems, and in building inputs. Facial recognition and forecasting have created new challenges and define computer-aided design trends and specifications for IoT systems. Figure 1 has shown the process of face recognition done during the IoT process.