By Anantha Krishnan M
Express News Service
Bangalore: Photographs and videos are two vital intelligence inputs that make the mission meaningful and menacing. With the advent of digital imagery, the option to process, enhance and disseminate the imagery intelligence opened up. Today, the trend is fast-pacing towards ubiquitous imaging, grabbing of videos rather than still images, and of creating large image collections.
The call of the hour hence is for specific technologies to process, store and exploit large number of images and videos. A comprehensive image and video processing solution that brings together the various processing capabilities that exist in the civilian space. Hence, the technology gets templated into three parts: a) ability to process the raw images and videos and convert them into higher quality or of lesser size; b) ability to create large collections which can be maintained and queried over military networks. [If repositories are not maintained and searchable, they are not useful in times of need.]; c) (iii) ability to automatically characterize images and videos, so as to allow object and action recognition of military interest for the volumes of images and videos likely to be acquired in the future.
Creation of such facilities requires a lot of technology development, integration and product design. There is an ongoing effort at the Centre for Artificial Intelligence and Robotics (CAIR) towards implementing, improvising and integrating the latest image and video processing methods to create a unified image and video application suite in the defence context.
In the area of computer vision, the CAIR has developed a comprehensive library for image and video processing, including techniques for image enhancement, manipulation, registration, geo-coding, multi-sensor image fusion, mosaicing and 3D reconstruction from multiple views. “The scientists have made inroads in areas like depth from stereo images, motion detection, tracking, progressive image and video transmission formats, super resolution-magnification, segmentation and steganography. In content characterization area, object detection and activity analysis, there is ongoing work using color, texture, key points, visual words, video shot segmentation, video synopsis and activity characterization,” sources said.
Prototype applications have been developed for Progressive Image Transmission, Content Based Image Retrieval, Image Fusion (including change detection) and Image-Map Display. In the area of visualization, the CAIR has developed the interactive 3D Terrain Flythrough software.
Currently, the computer vision team at the CAIR is engaged in a project to develop a number of applications for image and video processing in the net-centric operations (NCO) context. “These applications are designed to address the requirements of smart push and smart pull of images and videos to and from shared repositories across a network. The focus is on enabling preprocessing imagery, meta-data binding to imagery, tag-content-based imagery retrieval and exploitation of retrieved images and videos,” sources said.
AIMS OF NET-CENTRIC OPS
* Noise reductions and enhancement techniques.
* Magnification and super-resolution of images-video frames.
* Mosaicing of images-video frames.
* Fusion of images from different sensors to create composites.
* Processing videos from moving platforms to create stabilized videos with wide field of view and tracking of moving objects.
* Geo-registration and geo-calibration of images and videos to maps.
* 3D calibration and computation of 3D measurements from multiple images of target objects (buildings, bridges and other large objects)
* Automatic video summary generation based on activity periods and key frames to create short digests from long surveillance clips.
* Content-based automatic indexing and retrieval of images-videos.
* Progressive transmission formats for images and videos to cater for truncated transmission over tactical networks due to jamming.
The call of the hour hence is for specific technologies to process, store and exploit large number of images and videos. A comprehensive image and video processing solution that brings together the various processing capabilities that exist in the civilian space. Hence, the technology gets templated into three parts: a) ability to process the raw images and videos and convert them into higher quality or of lesser size; b) ability to create large collections which can be maintained and queried over military networks. [If repositories are not maintained and searchable, they are not useful in times of need.]; c) (iii) ability to automatically characterize images and videos, so as to allow object and action recognition of military interest for the volumes of images and videos likely to be acquired in the future.
Creation of such facilities requires a lot of technology development, integration and product design. There is an ongoing effort at the Centre for Artificial Intelligence and Robotics (CAIR) towards implementing, improvising and integrating the latest image and video processing methods to create a unified image and video application suite in the defence context.
In the area of computer vision, the CAIR has developed a comprehensive library for image and video processing, including techniques for image enhancement, manipulation, registration, geo-coding, multi-sensor image fusion, mosaicing and 3D reconstruction from multiple views. “The scientists have made inroads in areas like depth from stereo images, motion detection, tracking, progressive image and video transmission formats, super resolution-magnification, segmentation and steganography. In content characterization area, object detection and activity analysis, there is ongoing work using color, texture, key points, visual words, video shot segmentation, video synopsis and activity characterization,” sources said.
Prototype applications have been developed for Progressive Image Transmission, Content Based Image Retrieval, Image Fusion (including change detection) and Image-Map Display. In the area of visualization, the CAIR has developed the interactive 3D Terrain Flythrough software.
Currently, the computer vision team at the CAIR is engaged in a project to develop a number of applications for image and video processing in the net-centric operations (NCO) context. “These applications are designed to address the requirements of smart push and smart pull of images and videos to and from shared repositories across a network. The focus is on enabling preprocessing imagery, meta-data binding to imagery, tag-content-based imagery retrieval and exploitation of retrieved images and videos,” sources said.
AIMS OF NET-CENTRIC OPS
* Noise reductions and enhancement techniques.
* Magnification and super-resolution of images-video frames.
* Mosaicing of images-video frames.
* Fusion of images from different sensors to create composites.
* Processing videos from moving platforms to create stabilized videos with wide field of view and tracking of moving objects.
* Geo-registration and geo-calibration of images and videos to maps.
* 3D calibration and computation of 3D measurements from multiple images of target objects (buildings, bridges and other large objects)
* Automatic video summary generation based on activity periods and key frames to create short digests from long surveillance clips.
* Content-based automatic indexing and retrieval of images-videos.
* Progressive transmission formats for images and videos to cater for truncated transmission over tactical networks due to jamming.
Copyright@The New Indian Express
Part XI, Tomorrow: Artificial
intelligence inspired by nature