IMAPCAR | Automotive Electronics | Applications

Top	Why Image Recognition?	Why IMAPCAR?
How It Works	Features and Specifications	Applications

Any system using image recognition to enhance vehicle safety must be highly dependable and easy to mass produce. It must be based on innovative technology and have the flexibility to respond rapidly to technology changes.Until now, the real-time processing necessary for moving-image recognition has not been possible without bulky specialized hardware.

IMAPCAR changes all that. NEC has created a processor that combines top performance and low power consumption. IMAPCAR's cluster of 128 8-bit processors can parallel-process data at a speed of up to 100 GOPS* allowing real-time video processing. IMAPCAR is also energy efficient, consuming less than 2W of power with each video processing operation. IMAPCAR was designed to operate reliably under extreme environmental conditions, in temperatures ranging from -40ºC to 85ºC.Since all of IMAPCAR's programming is processor-based, alterations and upgrades are easily performed.

Today, almost all cars come with airbags, which have greatly reduced the number of vehicle fatalities. Nevertheless, the number of vehicle accidents overall is still increasing. At NEC Electronics, we hope that image-recognition technology will enable the creation of a "collision-proof car," and sharply reduce the number of traffic accidents. IMAPCAR is a major step forward in this direction.

Introducing image-recognition processing in mass-produced vehicles The best solution needs to take into account processing performance, development costs and power consumption.

*GOPS: Giga Operation Per Second

	Performance	Development costs	Power consumption
IMAPCAR parallel processor	Parallel processing via 128 8-bit processors results in gives IMAPCAR image processing performance second only to optimally designed specialized hardware. IMAPCAR processes multiple pixels simultaneously, so it has approximately 2 to 4 times (10 times at the application level) the image-processing performance of a high-performance general-purpose processor in the 2.4 GHz range.	IMAPCAR's software operations are processor-based and programmable, so the technology will support software upgrades due to advances in algorithms and peripheral devices. This allows shorter development time and lower costs.	Since IMAPCAR operates at a speed of 100 MHz, power consumption is low.
High-performance general-purpose processor	General-purpose processors have limited image-processing performance because they can process only one pixel at a time, which greatly limits their speed.	IMAPCAR is programmable and highly flexible. Development times are short and costs can be kept under control.	Since the operating frequency of general-purpose processors is several GHz, power consumption and heat-generation are high.
Core and dedicated circuits	Although the DSP core has strengths when it comes to sequential processing, it is not up to the task of parallel processing, which is so important in image processing. Even if dedicated wired circuits are used to increase the DSP's speed, it is still unable to reach the speed required for image-identification processing. The growing number of conditions and processing steps means that the scale of the dedicated circuits has to be increased. Moreover, the image-processing speed is not very high because each individual task is assigned to individual specialized circuits.	When the number of dedicated circuits is increased to handle the various processing conditions, development costs rise to the same high levels as those for specialized hardware.	In addition, their design is such that that increasing the number of circuits greatly raises power consumption levels.
Specialized hardware	When hardware is designed for a specific type of processing, its performance is optimized for the task.	The costs of keeping such a system up to date are considerably higher. The greater the number of different requirements for the system, the larger the chip size. With each advance in algorithms and peripheral devices, the system's hardware design has to be completely revised. In light of the steady advances that are taking place in the field of image recognition, this is a huge risk.	IMAPCAR features a specialized design and offers many cost advantages.

High-performance general-purpose image processing involves the high-speed processing of pixels one at a time by a single processor (dot image processing), a process that is inefficient in terms of both energy consumption and performance.

There are limits to the processing performance of even modern high-end, high-speed processors; they are simply not capable of real-time recognition and processing of video images. IMAPCAR uses the same kind of distributed parallel processing technology that is used in supercomputers. IMAPCAR divides up the series of pixels among its 128 individual microprocessors, and has them process the video stream simultaneously (linear image processing). In this way, IMAPCAR raises image processing far above the level attainable with power-hungry high-speed general-purpose processors, with only a fraction of the power consumption.

BACK