Many industries, from oil and gas to the power grid, are gaining a great deal of importance for real-time monitoring and surveillance. In recent years, surveillance applications have become increasingly demanding for high-quality real-time streaming, requiring advanced encoding and decoding techniques and computing-intensive image processing designs.
It takes expertise in a wide range of embedded technologies to design and implement a high-performance, real-time video streaming camera. This article aims to examine how the live streaming video of surveillance cameras is designed. A few streaming protocols besides camera architecture and video streaming are discussed, along with sensor selection and processing engines, requirements for system software as well as cloud integration.
Considerations for design
Image sensors
Pixels are discrete photodetector sites that comprise the image sensor of a camera system. A CCD or CMOS sensor is used for all cameras -- still, movie, industrial, and security photography. Sensors are responsible for converting light falling on lenses into electrical signals, which, in turn, are transferred by the camera's processor into digital signals.
Signal processing engines
Many SOMs and SBCs available on the market are based on leading GPUs, SOCs, and FPGAs, which meet the needs of the live streaming video of surveillance cameras. This large pool of processors may make it difficult to choose the right one. The camera system must be able to encode, decode, and process video in a way that meets the demands of the evolving video application. Also, the processors should support key camera parameters such as autofocus, auto exposure, image quality enhancement, pixel correction, and color interpolation.
With the advent of hardware accelerated encoding and streaming (GPU), image processing applications have been greatly enhanced, especially those related to surveillance and machine vision. For high-performance, real-time applications, GPUs provide more computing power than CPUs.
System software
Linux dominates the embedded world despite the emergence of many operating systems. Its stability and networking capabilities have made it the preferred embedded OS for years, particularly for connectivity and communications. Due to the numerous Linux vendors offering various kernel releases, developers can use Linux for HD Streaming Camera designs.
Algorithms
In a real-time, live streaming video of surveillance cameras, algorithms can be implemented in two layers. The first layer of algorithms focuses solely on processing images, while the second layer adds intelligence to those images.
Images captured by high-definition cameras are highly detailed. To make accurate decisions, it may be necessary to enhance further the data captured in order to make an effective analysis. For modern surveillance cameras, visual algorithms have become a minimum requirement, especially those that help improve image quality. A range of algorithms is commonly implemented in cameras, including autofocus, automatic exposure, histograms, color balancing, and focus bracketing.
Key performance and feature considerations
HD / 4K Video Streaming
The popularity and maturity of high-definition video streaming are growing. For aviation and security applications, a huge amount of data can pose major challenges to real-time HD video streaming. HD/4K video streaming faces many challenges, including transmitting large amounts of data with minimal latency to a remote location.
FPGAs, in particular, can be used to address this issue to a great extent. With an FPGA, developers are able to easily integrate complex video analytics algorithms into their devices. FPGAs can handle fast image processing and real-time HD video streaming thanks to their logic cells and embedded DSP blocks.
Video compression
For faster transmission to the live streaming video of surveillance cameras, raw video files are digitized, compressed, and packetized. The camera compresses a video using various techniques and algorithms, which the host PC can then reconstruct to its original resolution and format. Image compression standards, such as MJPEG, MPEG-4, H.263, H.264, H.265, and H.265+, can easily transfer images.
Video streaming protocols
Several video streaming protocols can be implemented in cameras to stream video over a network in real-time. In addition to the Real-Time Streaming Protocol (RTSP), which facilitates the efficient delivery of streaming multimedia over the network, there is also the Real-Time Transport Protocol (RTP), which transports multimedia streams over the network, as well as HTTP, transferring multimedia files over IP networks, including images, sound, and video. High-speed streaming cameras can benefit from the RTSP protocol, which combines with RTP and RTCP to provide low-latency streaming.
The form factor is small : the power is low
A great deal of pressure is being put on embedded product developers to reduce the size of their products while reducing their efficiency dramatically. Increasingly, the availability of low-power, small form-factor multicore SoCs, GPUs, and memory technologies are enabling designers to build cameras with small footprints and low power consumption.
Due to the nature of their deployment, industrial cameras, as well as surveillance cameras, must have low-power designs. Low energy dissipation and long battery life are requirements for such cameras designed for challenging environments.
Cloud integration
In the cloud, systems can be deployed at multiple locations, data can be stored in multiple locations, and advanced video analytics can be integrated, which enhances the overall system performance. Computer vision, artificial intelligence algorithms, and cloud computing technologies are all combined to deliver cutting-edge surveillance systems.
With AI algorithms, users are able to gather valuable insights from the data they are capturing. In contrast to edge computing devices, cloud computing may produce a slight delay in decision-making due to signal latency.
Conclusion
Real-time live streaming video of surveillance cameras is the need of the hour. As the embedded world expands, high-quality imaging plays a vital role in industrial robots, drones, surveillance, and medical applications. Using high-speed industrial interfaces and protocols, advanced cameras can capture and transfer large amounts of data. Developing a high-performance video streaming camera for surveillance applications requires understanding the fundamental components and exploring key design considerations. Using cutting-edge SOCs, the latest sensors, and a variety of image processing technologies and streaming platforms, one can create a real-time camera. Cameras must also meet environmental and temperature standards and have a properly designed mechanical design.
