Challenges in Implementing Smart Video Surveillance Systems — and the Advantage of Blackfin
Though tremendous advances have been made in the field of intelligent video surveillance, there is no method widely accepted as universally optimal. The inherent complexities in the research have caused a diversification of research measures and tools, limited applicability, and a lack of a standardized method aimed at carrying universally satisfying robustness, accuracy and speed.
Meanwhile, the networking of video surveillance, the demand for distributed processing, and the limitation placed on the cost, volume and power consumption in large-scale projects, have made powerful embedded processors the mainstream option for video analytics processing solutions.
The Blackfin processor, born from a joint development between ADI and Intel, has a differentiated Micro-Signal Architecture (MSA) which combines micro-controller capabilities with the traditional high-speed calculation capabilities of a DSP. MCU and DSP functionality is realized in the same core, with a single ‘converged’ instruction set. The development tools for Blackfin allow the creation and deployment of both control and signal processing code under a single tool-chain environment. Compared with a heterogeneous DSP + ARM chip architecture, Blackfin offers the advantage of a simplified software, as well as hardware, implementation. Blackfin is supported by over ten embedded operating systems, like ThreadX, Nucleus, uCOS-II, and uClinux, and thus provides a familiar software base to a variety of customers. Ideally positioned as a video processor, Blackfin has been specifically optimized for advanced data rate digital and media processing, and holds an extremely high performance-cost ratio. It also boasts high efficiency, delivering high performance at low power, making it suitable for IP camera products with small form factors.
Multiple DMA channels and flexibly configurable caches match well to the demand for the massive calculations and high data throughput found in video surveillance systems. The ten-stage pipeline makes the Blackfin capable of effective parallel instruction execution. Control instructions with zero overhead make the large amount of looping jump in the system consume no clocking cycles of processors. With these advantages, the running speed of a real decoder’s idct 4×4 algorithm has gained a 7× increase when the algorithm is running on a Blackfin processor.
Digital video data formats have certain inherent properties that signal processors can use to their advantage. In various chromatic spaces, 8 bits represent each component of the color data of a pixel. The four video mathematical calculation units and the video pixel instruction set of Blackfin use this 8 bit data width to allow parallel processing of multiple color components in a single instruction, greatly accelerating video operations. The video pixel processing instructions can accomplish, in a single cycle, eleven video pixel operations, such as addition, subtraction, mixed addition-subtraction, averaging or subtraction plus absolute-valuing, on four pairs of video data components. These operations find wide applications in various algorithms, such as motion estimation, loop filter and smart video analysis. The fundamental operations found in video analysis; bar charts statistics, median calculation, Sobel operation, and the expansion operations, may all utilize the MIN/MAX instructions of Blackfin to eliminate conditional jumps, further saving processor cycles. Blackfin can also support 13 vector operations for non-video data. By appropriately designing the data structure, instructions can be utilized in many stages (such as fore/background separation, threshold operation/updating) allowing video analytics algorithms become more efficient. Most of these increased efficiency instructions are capable of parallel execution, effectively doubling the Blackfin’s processing capability.
Analog Devices also simplifies the software development task by providing customers with a number of license-free and highly optimized audio and video encoder and decoder binaries. The users of Blackfin may review all the software module information at http://www.analog.com/processors/platforms/blackfinSoftwareModules.html .
NEXT: Intelligent Video Surveillance System Example
Page 3: next page
