Product Suite


ADAS Product Suite

Uurmi has developed the most advanced set of algorithms for a variety of ADAS applications such as Pedestrian Detection, Blind Spot Detection, Traffic Sign Detection, Collision Avoidance, Lane Detection, Backup Obstacle Detection and Occupant Sensing.


Uurmi licenses these algorithms to Automotive OEMs and Tier1 suppliers. In addition we also provide algorithm fine tuning and porting services for a variety of embedded platforms and processor systems.

Our algorithms take advantage of the technologies such as Biologically inspired visual attention and gist, 2D and 3D object recognition, Evolutionary optimization methods, Texture analysis, Shape-based recognition, Morphology and Image enhancement operations, Uncertainty handling in images using fuzzy sets, Tracking and fingerprinting algorithms, Segmentation of intensity and range images, Uncertainty modeling in data sets, Context-aware target recognition and Machine learning and training of classifiers Sample Applications.

Traffic Sign Recognition

Uurmi has developed solutions for Traffic Sign Detection. Our solutions use state of the art proven algorithms which can be ported on target hardware for real-time operation.

We have developed Traffic Sign detection where camera recognizes the standard characteristics of road signs in its line of sight; these are compared with road signs legend stored in the system and displayed if there is a match.

TSR Specifications

  • Chroma Based Segmentation
  • Sign Classification Engine
  • Recognize up to 15 different Traffic Signs with > 98% accuracy and < 5 False Alarm’s per hour
  • Can be extended to speed limit monitoring
  • Runs at 15 FPS on Exynos-5: ARM-A15 Dual Core + Mali-T604

Pedestrian Detection

Uurmi has developed solutions for Pedestrian detection. Our solutions use state of the art proven algorithms which can be ported on target hardware for real-time operation.

Uurmi’s unique approach to pedestrian detection lies in the use of monocular cameras only, using advanced pattern recognition and classifiers with image processing and optic flow analysis. Both static and moving pedestrians can be detected to a range of around 60m using 720p HD resolution imagers with >95% accuracy and at 10fps.

Pedestrian Detection Specifications

  • HOG based Classification
  • GPU for High Performance
  • Intelligent Scan Search
  • Recognize Pedestrians with 95% detection rate and 0.01% False Alarm Rate
  • Runs at 6 FPS on Exynos-5: ARM-A15 Dual Core + Mali-T604

Night Vision System

  • Vision-range enhancement and obstacle detection
  • Near-Infrared (NIR) system, PAL/NTSC composite output, VGA
  • OSRAM NIR LEDs, custom heat sink and illumination module
  • Range 100m, 30-60m alert zone, 60-100m warning zone
  • Detection Accuracy, >90% with < 2% false positives
  • Processing time 125ms for VGA image size
  • Detect obstacles using GPU enhanced code
  • Remains effective with oncoming head lights
  • Operate in urban, rural, and highways scenarios
  • Highlight free space in path of the host vehicle in red
  • Algorithm can be adapted and optimized on chosen hardware platform

Pedestrian Detection

Traffic Sign Detection


Lane Tracking

Night Vision