UZAY
Machine Vision and Pattern Recognition
Forensic Informatics
(Payload Interface Equipment)
Digital forensics involves uncovering valuable evidence through the meticulous examination and analysis of digital images, particularly in the field of image processing. This specialized branch of forensic science uses advanced techniques to detect, recover and analyze visual data from a variety of digital devices. By examining image metadata, detecting forgery or alteration, and enhancing image clarity, digital forensics experts can extract vital information for legal investigations, cybersecurity and law enforcement. The integration of image processing with digital forensics enables accurate and efficient analysis and plays an important role in solving crimes and verifying visual evidence.
Software Features
Features
Forensic software has been developed in the fields of Image/Video Positioning, Image Validation, Social Media Detection, Brand Model Detection, Image Forgery Analysis and Distance Estimation from Steryo Images. In all the analyzes performed, there is no need for any prior knowledge other than visual content. For example, in the Image/Video Localization Software, only the content information of a given query image or video is analyzed to determine from which coordinate it was taken. On the other hand, with Image Validation, database queries can be made with fingerprints extracted specific to the camera, regardless of brand, model and device type. Social Media Detection Software can be used to determine from which social media channel a query image was downloaded, and Brand Model Detection Software can be used to determine which brand and model family the query image belongs to. Image Forgery Analysis Software is used to determine whether malicious modifications such as copy-paste, montage or deletion have been made to the images. Finally, with the Distance Estimation Software, the absolute distances of the points in the scene are calculated using two or more images.
Smart Security Systems
(Payload Interface Equipment)
Software Features
Features
Security systems can be made intelligent with various artificial intelligence and pattern recognition methods, reducing the workload of operators and enabling more efficient use. These methods include crowd analysis, traffic analysis, license plate recognition, face recognition, object tracking and concept recognition, starting with background subtraction at the lower level and progressively more complex operations. Recognition of security-related concepts or event detection is more complex than the others as it aims at semantic level analysis and is still an important topic of academic research.
With the software developed by TÜBİTAK UZAY, automatic recognition of concepts over visual and auditory data and model-based learning over labeled data are provided. In the test phase, the conformity of the test data to the trained models is examined and the presence or absence of the searched concept is decided. The learned models can be updated with new data and the sensitivity of the system can also be adjusted. In this way, the risk of non-detection is minimized by taking false warnings into consideration.
BALISTICA®
(Payload Interface Equipment)
When firearms are used, they leave their own characteristic traces on the fired cartridges and casings. By comparing these marks on two different shells or casings, it is possible to determine whether these shells or casings were fired from the same weapon and to establish the relationships between the events.
BALİSTİKA® is a system that obtains 3D data of bullet and cartridge cases, stores this data in a database with accompanying metadata, and can store information about events, individuals, and firearms. It generates features by extracting attribute data from 3D data, compare them, and provide the probability of being a match. The system is able to distribute the workload of the comparison process among the high computation power cluster, provide computer-based comparison of bullet and cartridge cases and exchange data with other BALİSTİKA® labs in different regions.
Hive Entry Unit
Features
The Hive Entry Unit consists of high-tech equipment that enables the acquisition of 3D hive images. 3D hive images are obtained by processing 2D hive images. The hardware is controlled by the recording system software. With the help of a specially designed hive holder, images of hives of different diameter types can be taken. Recording of hive images is easily completed without requiring additional settings for the user.
Hardware Features
- Creation of high-resolution 3D hive images,
- Allowing the entry of barrels of different diameter types,
- Simultaneous visualization of the entire surface of the bucket,
- Specially designed shell holder,
- Since user intervention is not required in evidence image processing, images are taken at the same standard.
Core Input Unit
Features
The Core Entry Unit consists of the hardware that enables the acquisition of 3D core images. This hardware is also controlled by the recording system software like the Shell Entry Unit. It is ensured that 360-degree panoramic set images of the bullet core can be taken. With the help of a specially designed shell holder, it is possible to easily take images of deformed cores as well as images of cores of different diameter types.
Hardware Features
- Generation of high-resolution 3D bullet core images,
- 360 degree panoramic set images can be taken,
- Ability to take images of deformed nuclei,
- Simultaneous visualization of the entire surface of the bullet core,
- Since user intervention is not required in evidence image processing, images are taken at the same standard.
Semantic Video Analysis
(Payload Interface Equipment)
Software Features
Features
With the support of the Radio and Television Supreme Council (RTÜK), RTÜK SKAAS (2006-2008) and KaVTan (2008-2011) projects were carried out within TÜBİTAK UZAY to monitor, archive and analyze terrestrial broadcasts. In this context, various software programs such as Video Footage Analysis, Video-Text Reading, Video Program Analysis, Video Clip Capture, Keyword Capture and Radio Advertisement Analysis were developed to facilitate the analysis of experts monitoring terrestrial broadcasts:
Video Shooting Analysis: Software that automatically detects scene transitions within a video, displays each scene with a key frame, and enables quick video review.
Video-Text Extraction: Software that detects and records horizontal text overlaid on terrestrial broadcast content and provides quick access to video points where these texts appear through text query.
Video Program Analysis: Software that determines the time and duration of terrestrial broadcast commercials and identifies broadcasts that do not comply with RTÜK regulations.
Video Clip Retrieval: Software that searches a given query video clip in a previously created archive to determine which channels, when and how many times it has been broadcast.
Keyword Detection: Software that detects specified keywords in audio data independently of the speaker.
Radio Advertising Analysis: Software that automatically detects radio segments from audio data and identifies violations.
In addition, the ability to automatically recognize and track concepts, which are the abstract and general design of an object or thought in the mind, using visual and auditory data has been developed to automatically evaluate large-scale terrestrial broadcast data. In this context, a hierarchical concept ontology was created, five of which (violence, sexuality, illegal organization, human existence, nature) are at the top level, and the concepts trained with sample data were associated with each other according to this ontology.
