A5047089128- Advanced Neural Network Applications
- Industrial Vision Systems and Defect Detection
- Advanced Image and Video Retrieval Techniques
- Remote-Sensing Image Classification
- Video Surveillance and Tracking Methods
- Infrared Target Detection Methodologies
- Image Enhancement Techniques
- Advanced Image Fusion Techniques
- Oil Spill Detection and Mitigation
- Remote Sensing and Land Use
- User Authentication and Security Systems
- Smart Grid Security and Resilience
- Digital and Cyber Forensics
- Vehicle License Plate Recognition
- Spectroscopy and Chemometric Analyses
- Privacy-Preserving Technologies in Data
Institute of Space Science - INFLPR Subsidiary
2018
Solving the person re-identification problem involves making associations between same person’s appearances across disjoint camera views. Further, those have to be made on multiple surveillance cameras in order obtain a more efficient and powerful system. The becomes particularly challenging very crowded areas. This mainly happens for two reasons. First, visibility is reduced occlusions of people can occur. due congestion, as number possible matches increases, becoming achieve. Additional...
Pallet detection and tracking using computer vision is challenging due to the complexity of object its contents, lighting conditions, background clutter, occlusions in industrial areas. Using semantic segmentation, this paper aims detect pallets a logistics warehouse. The proposed method examines changes image segmentation from one frame next taking into account position stationary behavior newly introduced objects scene. results indicate that can despite contents. This demonstrates utility...
This paper presents an interface designed for the Darknet neural network, used by state of art YOLO (You Only Look Once) models. The object detection is still representing a challenge in field computer vision, and this has main purpose providing way to generate more easily new networks, order obtain desired system. Furthermore, through interface, it can be generated displayed feature maps at any point network. Therefore, presented paper, leads better understanding how network makes certain...
The paper presents an original neural network approach for region of interest detection and classification in multi-spectral satellite images. proposed method uses a sequence Pulse Coupled Neural Networks that identifies plausible regions interest. These are passed to dimension reduction algorithm, Principle Component Analysis, order generate the input data Support Vector Machine classifier, validates data. algorithm's parameters optimized using Genetic Algorithm. algorithm is designed...
Earth Observation (EO) through remote sensing satellites gathers reliable information about the environment. The evolution of satellite instrumentation leads to considerable data volumes which require on one hand, mathematically and statistically based algorithms extract valuable and, other increased computational power for processing. paper proposes image processing Synthetic Aperture Radar (SAR) images together with application-specific hardware architectures imagery. It is well known that...
Nowadays, the field of long-range thermal image processing is increasingly popular. Although this technology has several advantages in terms security and monitoring, there are not enough autonomous algorithms for these images. In paper, we aim to address two main problems field: lack databases shortage monitoring systems. Firstly, introducing a new proprietary database with Further, present detection algorithm, capable running on real-time This algorithm based general-purpose DNN (Deep...