Generally speaking, the orientation of an image can easily be recognized by human beings. On the other hand, recognizing the orientation of an image is one of the challenges facing digital image processing. Several attempts have been made in this research area. However, the optimal solution still needs to be determined. For such a solution, the main issue is to obtain the precise angle for possible rotation. Numerous attempts have been made to address this issue; some of them mainly focus on coarse angles, whereas others have applied fuzzy logic in order to determine more precise angles. In this paper, a two-stage method is introduced in which a coarse angle estimation is achieved through the use of the Convolutional Neural Network (CNN) approach, and a more precise angle is acquired via fuzzy logic. An extensive evaluation of the proposed method is carried out on different public datasets. The results indicate an outstanding level of performance in terms of optimizing image orientation. INDEX TERMS Orientation of an image, precise angle estimation, convolutional neural network (CNN), fuzzy logic. A. OBTAINING AN ESTIMATE OF THE COARSE ANGLE (0 • , 90 • , 180 • , AND 270 •)

Commercial airlines and the passengers suffer from flight delay. Flight delay causes huge loss for the airlines and unsatisfied passengers. The researchers attempt to solve this problem through prediction extensively by machine learning approach and data mining tools. Accurate and robust performance is still to get through existing models. Our proposed hybrid approach is intended to use the power of machine learning as data mining tool and to predict the delay using classification algorithm of deep learning. An extensive evaluation of the proposed method is carried out by comparing the performance by using two data sets: one is local and the other is benchmark from Kaggle to obtain the best performing classifier. Three predictive models were applied on the datasets: logistic regression, decision tree and the proposed approach. The result shows that the proposed method performed well as comparing to the existing state-ofthe art.

No abstract

Visual Object Tracking (VOT) is the most salient and an ongoing exploration field amongst the several disciplines of computer vision. The importance of this technology is due to the extensive range of applications such as robot navigation, human computer interaction, video surveillance, etc. The process of object tracking involves segmenting areas of a video scene and tracking its position, motion and occlusion. However, problems can appear during tracking on account of multiple issues including camera motion, object-to-object and object-to-scene occlusions, nonrigid structures, object and scene changes in patterns and appearance and abrupt object movement. The aim of this paper is to examine, analyze and provide a shortlist of the most ubiquitous object tracking techniques. This accomplish by providing a comprehensive review of the tracking process which involve object detection methods, object representation and features selection and object tracking over multiple frames. Object tracking methods are compared whilst elaborating upon the advantages and limitations.

(rule table).