Technology Overview

Introduction

Video surveillance systems are commercially available since early 50s. A major part of the products are dedicated to capture, transfer and record video streams to be evaluated by human. Due to the latest advances in computer vision, video surveillance is now getting more and more unsupervised and its functionality is being extended to automatic recognition of situations and extraction of information out of life video sequences.

Image processing is still counted as an costly technology, however the gap between this technology and other measurement technologies is getting smaller. Beside the dimensions of such systems, the reliability of applied algorithms was presenting for many years the major hurdle in many application fields. The new trends of downsizing of hardware components and the latest improvements in the applicable algorithms have yielded more convenient and low priced embedded video and image surveillance systems. This was a crucial benefit for economical justification of such visual sensor systems compared to other sensor systems and brings them in many new application domains as a new option. Possessing a high accuracy and reliability in many application fields make such visual sensors to a serious competitor of other sensors. Other excellences of vision sensors are their flexibility by getting more functionality as a multisensory device just by running divers updatable softwares with the same hardware.

On the Fly Image Processing (OTFIP™)

HaViSys is the technology provider of On the Fly Image Processing (OTFIP™). OTFIP™ is a software technology  aiming to reduce the resource consumption of embedded image processing. In this method the images are streamed through the processing stages without being saved in the memory. Processing stages abstract image contents and extract the results while the image is being transmitted. Final results are ready to report as soon as the image is completely streamed.  Following graphic shows a schematic view of a comparison between OTFIP™ and conventional image processing technology:

Conventional image processing requires a high access-rate to the memory mobined with the huge amount of data being transferred back and forth, this forms the main bottleneck for the execution of algorithms in this method. Caching is usually used to cover the latency of the memory units, however this method fails due to the random access to large areas pursued by conventional image processing algorithms. It is worth noting that the RAM access latency is 10 to 20 times higher than the program execution.

OTFIP™ technology addresses this problem by sequentially processing image data so that all storage activities can be performed locally using tightly coupled memory of the processing device. HaViSys offers its know how as a comprehensive library for OTFIP™ algorithms. The optimized hardware platforms for these software are also developed and manufactured by HaViSys and represent our vision sensors.

Advantages of OTFIP™ Technology

OTFIP™ technology can simplify the hardware of vision sensors and brings the following benefits:

  • COG: Lowers fabrication costs by omitting the costly components
  • Reliability: Higher reliability by less usage of susceptible hardware and software components
  • Energy: Lower energy consumption by using low power hardware components and slimmed data usage
  • Bandwdth: Narrower communication band by producing smaller data volumes after an on-board processing operation, this facilitates the application of low cost communication gateways e.g. GPRS.
  • Scalability: Sensor Scalability of OTFIP devices as a low cost system is possible for use cases in large area surveillance coverage e.g. in Smart city applications.