Yazar "Sonugur, Guray" seçeneğine göre listele

Listeleniyor 1 - 2 / 2
  • Küçük Resim Yok
    Öğe
    A NEW GRADIENT-BASED FEATURE EXTRACTION METHOD FOR REAL-TIME DETECTION OF MOVING OBJECTS USING STEREO CAMERAS
    (Publ House Bulgarian Acad Sci, 2022) Sonugur, Guray; Gokce, Baris
    In this study, a gradient-based feature extraction method has been developed that can be used to detect moving objects in real-time applications such as unmanned ground or air vehicles. Feature extraction methods should produce fast results in real-time applications, as results need to be obtained between successive frames of video sequences within a limited time. For this reason, various sized image blocks were used in the developed method. The arithmetic mean (AM), geometric mean (GM), median (MD), and local contrast (LC) methods were used to calculate block intensities. In the stereo video stream, depth maps were also divided into blocks along with successive frames' R, G, and B channels. A novel feature extraction method was developed by calculating gradient-based relationships between adjacent blocks around the centre block. In experimental studies, the features extracted from stereo video frames using the proposed method were compared with Surf, Fast and Brisk methods according to their quantity, accuracy, and processing times, and more successful results were obtained. In addition, the moving object detection performance of the method was tested in real-time using an Unmanned Ground Vehicle.
  • Küçük Resim Yok
    Öğe
    Recognition of dynamic objects from UGVs using Interconnected Neural network-based Computer Vision system
    (Taylor & Francis Ltd, 2022) Gokce, Baris; Sonugur, Guray
    In this study, moving object recognition is performed by using images from a camera mounted on an unmanned ground vehicle. A GPS coordinate-based algorithm has been developed to obtain moving object silhouettes. In order to classify these silhouettes, an interconnected artificial neural network (ICANN) architecture consisting of two stages has been developed. The method consists of two phases. In the first phase, real-time images are converted to binary images at the end of the GPS-assisted image registration process. Then, the silhouettes are extracted from the background of the images using connected component labelling. In the second phase, two interconnected neural networks are used. The first neural network classifies silhouettes as objects or noise. The second neural network divides objects into seven subclasses as pedestrians, potholes, cars, etc. Compared to CNN-based techniques, a simpler NN architecture was employed in the research, and better accuracy rates were achieved with fewer samples. Another contribution of the research is simultaneous localization and mapping (SLAM) applications can be performed in non-GPS environments using pre-recorded images containing GPS information. In experimental studies, maximum success rates of 96,1% in object classification were obtained. The results obtained were compared to YOLO, the recently popular algorithm for object recognition.