Eyup Gedikli

Background Estimation Based People Detection and Tracking for Video Surveillance

This paper presents a real-time background estimation and maintenance based people tracking technique in an indoor and an outdoor environments for visual surveillance system. In order to detect foreground objects, first, background scene model is statistically learned using the redundancy of the pixel intensity values during learning stage, even the background is not completely stationary. A background maintenance model is also proposed for preventing some kind of falsies, such as, illumination changes, or physical changes. And then for people detection, candidate foreground regions are detected using thresholding, noise cleaning and their boundaries extracted using morphological filters. From these, a body posture is estimated depending on skeleton of the regions. Finally, the trajectory of the people in motion is implemented for analyzing the people actions tracked in the video sequences. Experimental results demonstrate robustness and real-time performance of the algorithm.

A novel approach on silhouette based human motion analysis for gait recognition

This paper presents a novel view independent approach on silhouette based human motion analysis for gait recognition applications. Spatio-temporal 1-D signals based on the differences between the outer of binarized silhouette of a motion object and a bounding box placed around silhouette are chosen as the basic image features called the distance vectors. The distance vectors are extracted using four view directions to silhouette. Gait cycle estimation and motion analysis are then performed by using normalized correlation on the distance vectors. Initial experiments for human identification are finally presented. Experimental results on the different test image sequences demonstrate that the proposed algorithm has an encouraging performance with relatively robust, low computational cost, and recognition rate for gait-based human identification.

Gait Recognition by Applying Multiple Projections and Kernel PCA

Recognizing people by gait has a unique advantage over other biometrics: it has potential for use at a distance when other biometrics might be at too low a resolution, or might be obscured. In this paper, an improved method for gait recognition is proposed. The proposed work introduces a nonlinear machine learning method, kernel Principal Component Analysis (KPCA), to extract gait features from silhouettes for individual recognition. Binarized silhouette of a motion object is first represented by four 1-D signals which are the basic image features called the distance vectors. The distance vectors are differences between the bounding box and silhouette, and extracted using four projections to silhouette. Classic linear feature extraction approaches, such as PCA, LDA, and FLDA, only take the 2-order statistics among gait patterns into account, and are not sensitive to higher order statistics of data. Therefore, KPCA is used to extract higher order relations among gait patterns for future recognition. Fast Fourier Transform (FFT) is employed as a preprocessing step to achieve translation invariant on the gait patterns accumulated from silhouette sequences which are extracted from the subjects walk in different speed and/or different time. The experiments are carried out on the CMU and the USF gait databases and presented based on the different training gait cycles. Finally, the performance of the proposed algorithm is comparatively illustrated to take into consideration the published gait recognition approaches.

Gait recognition using view distance vectors

This paper presents a new approach for human identification at a distance using gait recognition. Binarized silhouette of a motion object is represented by 1-D signals which are the basic image features called the distance vectors. The distance vectors are differences between the bounding box and silhouette, and extracted using four view directions to silhouette. Based on normalized correlation on the distance vectors, gait cycle estimation is first performed to extract the gait cycle. Second, eigenspace transformation based on PCA is applied to time-varying distance vectors and then Mahalanobis and normalized Euclidean distances based supervised pattern classification is finally performed in the lower-dimensional eigenspace for human identification. Experimental results on two main database demonstrate that the right person in top two matches 100% of the times for the cases where training and testing sets corresponds to the walking styles for data set of 25 people, and other data set of 22 people.

Silhouette Based Gait Recognition

This paper presents a approach for gait recognition based on binarized silhouette of a motion object which is represented by distance vectors. The distance vectors are differences between the bounding box and silhouette, and extracted using four projections to silhouette. First, gait cycle estimation is performed based on normalized correlation on the distance vectors. Gait patterns are then extracted by using distance vectors for each projection independently. Then gait patterns are normalized according to dimensions of bounding box and gait cycle. Second, PCA based eigenspace transform is applied to gait patterns and Euclidean distances based supervised pattern classification is finally performed in the lower-dimensional eigenspace for human identification. Experimental results on four databases (CMU MoBo, SOTON, USF, NLPR) show that the proposed approach achieves highly competitive performance with respect to the published gait recognition approaches.

Microscopic image segmentation based on firefly algorithm for detection of tuberculosis bacteria

One third of the world is infected with tuberculosis disease. The disease is diagnosed visually by laboratory technicians. In the microscopy diagnosis with hand-eye control, misdiagnosis rate is quite high. In microscopic imaging, by using computer aided automatic diagnosis methods, the disease is true diagnosed. The robustness of the automatic diagnosis methods depends on accurate segmentation of microscopic images. Image segmentation methods produce a special solution for several problems. In this study, Firefly algorithm based on swarm intelligence as a novel approach in microscopic imaging is proposed to segment images. In the proposed approach, an optimum threshold value in gray-level microscopic images is determined with proposed entropy based Firefly algorithm. Microscopic images are converted to binary format by using obtained optimum threshold value. Segmentation results are compared with expert-guided segmentation results. The performance ratio of segmentation is 96% obtained by using Firefly algorithm based on swarm intelligence.

Online identity verification system based on palmprint

A biometric verification system based on palmprint recognition works at the real enviroments is presented in this work. This system initially detects all objects entered in the view of a CCD camera which is settled into hand placement platform. The detected objects are classified as hand objects or not by applying the processes on the image sequences acquired from the camera. In the case of hand object, a palm area as region of interest (ROI) is then selected from the hand regions. The resolution and gray levels of the ROI are also normalized to achieve higher accuracy. In the pattern recognition stage, a gabor-palm image is first created by applying Gabor-based discrete transform, then two different feature extraction approaches (both PCA and KPCA) and two classifications (NN based WED, SVM) methods are simultaneously applied onto the gabor-palmprint images, respectively. In this system, the entrance of a PC room is controlled by authorising of the users whose are successfully verified. The enrollment and verification processes in the presented system are fully performed automatically. The experimental results performed on the real environments show that the proposed biometric system can be easily employed with highly reliable performances in the real applications.

PSO and SURF based digital image forgery detection

Digital images have become very important in our daily lives and some other important areas such as medicine, journalism and it can be also used as forensic evidence. However, the simplicity of using digital images with freely available software tools makes the authenticity of images questionable. The most common image forgery type is copy move forgery because it can be done easily but the detection of this type of forgery is hard. Various approaches are proposed in literature to detection of copy move forgery, but lots of them is not satisfy result especially smooth regions are used to hide objects. And lots of works use experience parameters values so sometimes they cannot detect forgery operations. To solve these problems we proposed an optimized keypoint based copy move forgery detection methods based on Speeded-Up Robust Features (SURF) algorithm and Particle Swarm Optimization (PSO). Experimental results show that the proposed method has good performance even under post processing and preprocessing attacks (such as blurring, noise addition, rotation, JPEG compression).

Elevated Plus Maze on Based Computer Vision

This paper presents a novel method for rat detection and tracking in a platform known as elevated-plus maze and for recording the rat movements as a time elapsed in specific regions in real time surveillance systems. First, the location of plus maze platform is automatically determined by using hough transform. Rats in the platform are detected by applying otsu based automatic thresholding. Otsu-based automatic thresholding is applied to discriminate the pixels as foreground and bacground pixels. The foreground pixels are then grouped by using a bounded box. From now on, for every new frame , active subject has been followed in platform that is seperated regions and elapsed times in regions are recorded. Plus maze system we take 25 frames per second is present more right results than other available systems. We applied this approach on video records that are taken from KTU-Physiology Lab.

Barnes Maze Based on Computer Vision and Learning

In this study, motion detection and tracking for real time system is presented. Through the low pass operations, the system works efficiently in real time. In the study, the Barnes Maze test mechanism is automatically learned by Hough transform. Background subtraction algorithms for object detection and estimation approaches based on color, shape and position for tracking are used. Since the desired results are related to the object organs, silhouette analysis is also used. The system observes the experiment mechanism. To detect the target (e.g. cheese), rat motions in the platform are tracked using camera vision system and then the motion positions in 2-dimensional are recorded. These data can be evaluated physically and psychologically. In this study making the learning model of the object from its behaviours is also the future work. For this purpose Markov processes could be used