Image processing, Machine vision and video analytics

The world around us is an overloaded digital world, covered by large number of multi-mega-pixel sensors, generating a vast amount of bits of data.

This data overload creates great challenges at the human operator level, as well as on an operational network level – immense difficulty to process / transfer such large quantity of data generated by the sensors, limiting our capability to put to good use all the data available to us, possibly ending in missing important information. Therefore, there is a great need to perform “smart” and automated decisions and data-reduction at the camera / payload level due to communication bandwidth bottlenecks. Modern EO systems may include a wide range of Machine Vision and Video Analytics capabilities which enable to automatically detect and classify targets of interests (pre-defined by the operator). In such case, the operator will receive and have to handle only a limited amount of efficiently targeted/concentrated data, meaning a significant reduction of operator overload and increase to mission efficiency.

High quality of AI neural network performance level is very much based on the following elements, which are defined as core features in many of CONTROP systems:

• High quality image – image processing and stabilization level
• Number of pixels on the target of interest – long focal length and high resolution
• High quality of the video analytic structure and algorithms

All CONTROP systems include advanced video processing algorithms and video analytics at the Camera / System Level:

• Raw video image processing algorithms – Non-uniformity corrections, Global and Local AGC, image enhancement and sharpness, anti-turbulence function. These functions are an essential prerequisite for the generation of an optimal image for the operator and / or the machine vision AI module, and by that increasing the automatic classification capabilities.
• Automatic Target detection and classification algorithms
  • Automatic Target Detection based on motion detection
  • Automatic Target Classification / Recognition (ATR) based on real time AI neural network.
  • Automatic target Detection based on unique algorithms and deep learning capabilities
• Additional video algorithms for operator aid:
  • Automatic Video Tracker (ATR) – high performance video tracker designed specifically for Air-Ground-Naval application. The algorithms include automatic target engagement, automatic platform maneuvering compensation, offset tracking and prediction algorithm.

A quality tracker is especially essential in target designation and/or firing missions.

• Video-Based navigation capabilities while operating within a GPS jammed/denied arena.

All the above-mentioned features are essential and important for automated and efficient scene control, becoming an operator’s vital tools for real-time mission performance and decision making.