Three major technologies of digital surveillance intelligent video analysis

In China's surveillance market, digital video recorders (DVRs) once occupied half of the video surveillance market. However, with the development of network surveillance, network cameras (IPCs) and network video recorders (NVRs) are also catching up. Among them, video intelligent analysis is becoming a hot spot in digital video surveillance, and DSP has become the driving force of this technology. Powerful SoC systems are also one of the driving forces of intelligent analysis technology.

　　From the perspective of technology development trends, network video surveillance will gradually become the mainstream mode; this is an unquestionable trend. With the development of network surveillance, intelligent video analysis has become the next important direction for the development of IP video surveillance. However, for most chip manufacturers and system manufacturers, as IP video surveillance gradually develops towards high-definition and intelligent directions, it brings them not only huge market opportunities but also new challenges. The main technical trends in designing digital video surveillance systems are the need to achieve high image quality (1080p resolution), video analysis, intelligent management, IP, H.264 video compression standards, higher image signal processing clarity, device functions for applications and matching project algorithm functions. These new trends require developers to face challenges in low power consumption, low cost and environmental protection.

　　Powerful SoC system

　　TI's digital signal processor (DSP) has strongly promoted the development of digital video surveillance technology, among which the milestone is the TMS320DM642 general-purpose digital media processor. DavinciTM, a digital media processing single-chip system launched by TI, provides a powerful platform for digital video surveillance. Davinci's various series of platforms can fully cover all application directions of digital video surveillance, including two main development directions: ARM as the main control processor and DSP as the main processor. Their common feature is that they can support multiple video compression formats, including H.264/MPEG-4, etc. All DSP cores have the ability to support VCA.

　　At present, the solution based on DSPSoC occupies the mainstream position in the monitoring market. At the same time, some FPGA manufacturers hope to show their advantages in differentiation and customization. By combining FPGA devices and MPU/DSP for monitoring applications, designers can benefit from the inherent flexibility of programmable coprocessors, and can also use FPGA to focus on developing algorithms to fully utilize the processing power of MPU/DSP. Some manufacturers have adopted DSP FPGA in their multi-channel video monitoring system hardware platform, using TI's TMS320DM6437, which has only one digital video input interface. In order to connect 8-channel video decoders, it is necessary to add coupling logic between the video decoder and the video input interface of DSP, and programmable FPGA can easily realize the coupling logic between the two.

　　In order to meet the needs of the future digital video surveillance market, greater flexibility is necessary. Video compression technology will develop towards intelligent encoding and decoding, traditional monitoring will still have to develop towards real-time intelligent analysis, and optical systems will be more closely integrated with digital electronic systems.

　　DSP processor improves intelligent analysis and processing capabilities

　　After several generations of technological evolution, video surveillance systems are moving towards networking (IP) and intelligence. In addition, the coexistence of analog and digital surveillance systems will also be a feature of the Chinese security market in the next few years. As a mid-to-high-end solution, DSP is currently the mainstream processor in the Chinese surveillance market. ASICSoC is a latecomer and has a certain impact on the market. In addition, FPGA is also trying to enter the market, but there are no obvious successful examples yet. Since there is no standard algorithm for intelligent analysis and processing at this stage, the MIPS requirements of DSP will be greatly different when implementing intelligent video analysis and processing based on different compression formats, so algorithm optimization is very critical. Due to its flexibility and high performance, DSP is better at intelligent analysis and processing.

　　The ideal DSP processor for image processing applications must have the following characteristics: strong core processing capability; instruction set specifically for image processing; low-power hardware architecture that facilitates large amounts of data transmission; high integration; rich software module library; and powerful development tools. For video surveillance systems, the most suitable control and processing should be a convergent processor that combines control functions and digital signal processing functions.

　　Most of the early video surveillance products in China used DSP. In multi-channel, high-resolution video encoding and decoding applications, it is difficult for DSP software to meet system requirements. FPGA uses hardware acceleration to perform multi-channel compression in real time. The advantage of FPGA is that it is programmable, which is very suitable for application designs that require high flexibility and provide differentiation. At present, video surveillance products using FPGA chips have been mass-produced.

　　H.264 video codec technology

　　In the history of video surveillance, there have been many video compression standards. The standards defined by the International Telecommunication Union (ITU) for low-bit-rate video telephony include H.261/2/3; the International Organization for Standardization (ISO) defines the MPEG standards for motion image compression for consumer applications, including MPEG1/2/4; the H.264 codec technology is defined by ITU and ISO in cooperation with target applications (Joint Video Team JVT), and some standards still exist today. Currently, the three main standards in use are M-JPEG, MPEG4 and H.264. We compare these standards through the following points.

　　Implementation cost: including the complexity of algorithm implementation and the degree of resource occupation, the difficulty and cost of technology acquisition, and the cost ultimately reflected in the product;

　　Coding performance: mainly reflected in the size of the compressed bitstream under the same subjective and objective visual quality, which is very critical for network transmission and storage;

　　Network support: The quality of network support mainly depends on the fault tolerance of network packet loss and errors and the adaptive ability of the supported network;

　　Application support scope: Since the application requirements in the field of video surveillance are ever-changing, the ability of a platform to support different requirements is also very important;

　　Future potential: From a development perspective, the potential of a standard is reflected in how widely it is accepted by the industry in the future, how smoothly it is transformed into related applications, and how well it supports interconnection and interoperability.

　　In video surveillance products, the encoder is always the focus of the solution. A suitable encoder solution can often determine the success or failure of a product. The H.264 encoding solution is becoming the mainstream of the market. It has been widely recognized by the market for its excellent encoding performance and good network adaptability. It can not only support a wide range of application environments, but also meet the needs of different resolutions, different bit rates, and different transmission and storage environments.

　　H.264 is an advanced video processing technology. It is generally recognized by industry insiders that under the same video playback quality conditions, H.264 saves nearly 40% of transmission bandwidth or hard disk storage space compared to MPEG4, which undoubtedly plays an important role in the promotion and application of remote monitoring. However, the coding efficiency of H.264 comes at the cost of its much more complex calculation than MPEG4. It uses the most advanced inter-frame prediction mode, including complex motion estimation, 1/2 and 1/4 pixel prediction; the intra-frame prediction mode is much more advanced than MPEG4, including up to 13 intra-frame prediction modes; H.264 introduces a new in-loop filtering technology, which has made a huge contribution to the substantial improvement of video playback quality.

　　However, the application of new technologies requires a lot of computing and processing, and also puts forward new requirements for the video codec processing platform. For software-based video codecs, such as codecs implemented by general-purpose processors such as DSP, due to the limitation of processing power, many of the most advanced technologies of H.264 (such as intra-frame and inter-frame coding prediction modes, loop filtering, etc.) can only be abandoned. As a result, its encoding effect returns to the level of MPEG4. What's more serious is that the code stream output by such encoders does not meet the H.264 standard, and all general decoders cannot form an open system with it. If a non-standard encoder is selected, users can only limit the selection of non-standard decoders that match it, thus becoming a closed system.