• Intelligence

    Surveillance and

    Reconnaissance Suite

ISR Suite

Uurmi have custom solution for bio-inspired saliency, attention, and search algorithms, neural·algorithms for object recognition, activity recognition, context extraction, multi-sensor fusion, and rea soning engine for exploitation. Our solutions use state of the art proven algorithms which can be easily ported on to target hardware for real-time operation in tactical, operational, and forensic applications.

Persistent Surveillance

  • Reliable detection of targets using and fusing visible, Radar, SAR, Lidar, Infrared sensors
  • Pre-processing algorithms that rectify fog, haze, rain, jitter, and atmospheric turbulence effects
  • Robust maritime surveillance in visual and SAR domain, texture recognition in SAR imagery
  • Can fuse multiple views and sensors into a synthetic vision panorama for high-level exploitation
  • Can do geo-registration, KLV tagging, and onboard H.264 streaming from low SWAP platforms
  • Can Port Algorithms to Nvidia Tegra, Freescale i.MX, ARM, MIPS, Tensilica, Marvell ARMADA series, TI, and Toshiba Visconti Platforms

Multi Sensor Fusion & Image Exploitation
Target Detection and Tracking

  • Target detection & tracking in following scenarios
  • - Fog/Snow, Atmospheric Turbulence and Dynamic Background
  • Randomize temporal diffusion of pixel based background estimation with target detection based on statistical analysis of foreground blobs
  • Ability to handle camera vibrations/shake, sensor noise and scene change scenarios
  • Runs "20fps VGA size on DSP/FPGA"

Sample results of recognition of different types of objects, humans, vehicles, signs, etc

Unique Attributes
  • Automatic selection offeatures and build ofthe classifier cascade using a cla ssification engine
  • Data fusion suite that uses many algorithms
  • locate and Recognize objects using fa st search methods and statistical classifiers
  • Use of multiple features including texture, shape, appearance, motion, flow, wavelength, and shadows forrobust recognition

Sensor Fusion across multiple bands of the spectrum including visible, IR, lIDAR, SAR, etc. Fusion include PCA-based approaches, weighted averaging, highpass filtering, pyramid tran sform, and wavelet-based.

Automatic Terrain Mapping and classification using LIDAR

  • Robust classification of point cloud data for situational awareness
  • Real-time on-board processing of lIDAR data
  • Recognition of15 different types of terrains and object types

Image Enhancement
Fog/Haze Detection

  • Image processing based in the environment and applies correction to see through fog
  • Spatial diffusion filtering based approach with local contrast enhancement to remove fog/haze
  • Fog correction range is "150 meter
  • Runs -9fps VGA size on ARM- A15/A7/T604 Mali
  • Fog correction for images captured from un-calibrated camera systems
  • Fog correction with scene sharpness preservation (no extra blurriness) Capable of fog / non-fog region separation

Low Light Image Enhancement
  • Low light correction using Inverse spatial filtering with local self-similarity
  • Sensor independent real-time low light enhancement
  • Ability to recover true colors of the scene
  • Low light and non-low light region detection
  • Ability to handle camera vibrations/ shake
  • Smooth, realistic low light corrected scene generation without patchy distortions
  • Ability to handle sensor noise generated in extreme dark conditions
  • Runs - 1Ofps VGA size on ARM- A15/A7/T604 Mali

Burst Mode Image Enhancement
  • Camera captures continuous shots and provides bunch of images with in a second
  • The information from multiple images is embedded into high resolution grid to get higher zoom level
  • Reconstruct distorted scene or objects and estimate missing information
  • Up to 5X zoom is possible


Target Detection and Tracking


Multi Sensor Image Fusion

Automatic Terrain Mapping and Classification using LIDAR

Fog/Haze Detection

Low Light Image Enhancement

Burst Mode Image Enhancement
Burst Mode

Burst Mode