In-Flight Tests of Intruder Detection Vision System.

In the near future, the integration of manned and unmanned aerial vehicles into the common airspace will proceed. The changes taking place mean that the safety of light aircraft, ultralight aircraft and unmanned air vehicles (UAV) will become an increasing problem. The IDAAS project (Intruder Detection And collision Avoidance System) meets the new challenges as it aims to produce technically advanced detection and collision avoidance systems for light and unmanned aerial vehicles. The work discusses selected elements of research and practical tests of the intruder detection vision system, which is part the of IDAAS project. At the outset, the current formal requirements related to the necessity of installing anticollision systems on aircraft are presented. The concept of the IDAAS system and the structure of algorithms related to image processing are also discussed. The main part of the work presents the methodology developed for the needs of dedicated flight tests, its implementation and the results obtained. The initial tests of the IDAAS system carried out on an ultralight aircraft generally indicate the possibility of the effective detection of intruders in the airspace with the use of vision methods, although they also indicated the existence of conditions in which this detection may prove difficult or even impossible.

A Vision-Based Method for Determining Aircraft State during Spin Recovery.

This article proposes a vision-based method of determining in which of the three states, defined in the spin recovery process, is an aircraft. The correct identification of this state is necessary to make the right decisions during the spin recovery maneuver. The proposed solution employs a keypoints displacements analysis in consecutive frames taken from the on-board camera. The idea of voting on the temporary location of the rotation axis and dominant displacement direction was used. The decision about the state is made based on a proposed set of rules employing the histogram spread measure. To validate the method, experiments on flight simulator videos, recorded at varying altitudes and in different lighting, background, and visibility conditions, were carried out. For the selected conditions, the first flight tests were also performed. Qualitative and quantitative assessments were conducted using a multimedia data annotation tool and the Jaccard index, respectively. The proposed approach could be the basis for creating a solution supporting the pilot in the process of aircraft spin recovery and, in the future, the development of an autonomous method.

Recognition of Signed Expressions in an Experimental System Supporting Deaf Clients in the City Office.

The paper addresses the recognition of dynamic Polish Sign Language expressions in an experimental system supporting deaf people in an office when applying for an ID card. A method of processing a continuous stream of RGB-D data and a feature vector are proposed. The classification is carried out using the k-nearest neighbors algorithm with dynamic time warping, hidden Markov models, and bidirectional long short-term memory. The leave-one-subject-out protocol is used for the dataset containing 121 Polish Sign Language sentences performed five times by four deaf people. A data augmentation method is also proposed and tested. Preliminary observations and conclusions from the use of the system in a laboratory, as well as in real conditions with an experimental installation in the Office of Civil Affairs are given.

Data Augmentation with Suboptimal Warping for Time-Series Classification.

In this paper, a novel data augmentation method for time-series classification is proposed. In the introduced method, a new time-series is obtained in warped space between suboptimally aligned input examples of different lengths. Specifically, the alignment is carried out constraining the warping path and reducing its flexibility. It is shown that the resultant synthetic time-series can form new class boundaries and enrich the training dataset. In this work, the comparative evaluation of the proposed augmentation method against related techniques on representative multivariate time-series datasets is presented. The performance of methods is examined using the nearest neighbor classifier with the dynamic time warping (NN-DTW), LogDet divergence-based metric learning with triplet constraints (LDMLT), and the recently introduced time-series cluster kernel (NN-TCK). The impact of the augmentation on the classification performance is investigated, taking into account entire datasets and cases with a small number of training examples. The extensive evaluation reveals that the introduced method outperforms related augmentation algorithms in terms of the obtained classification accuracy.

Recognition of Fingerspelling Sequences in Polish Sign Language Using Point Clouds Obtained from Depth Images.

The paper presents a method for recognizing sequences of static letters of the Polish finger alphabet using the point cloud descriptors: viewpoint feature histogram, eigenvalues-based descriptors, ensemble of shape functions, and global radius-based surface descriptor. Each sequence is understood as quick highly coarticulated motions, and the classification is performed by networks of hidden Markov models trained by transitions between postures corresponding to particular letters. Three kinds of the left-to-right Markov models of the transitions, two networks of the transition models-independent and dependent on a dictionary-as well as various combinations of point cloud descriptors are examined on a publicly available dataset of 4200 executions (registered as depth map sequences) prepared by the authors. The hand shape representation proposed in our method can also be applied for recognition of hand postures in single frames. We confirmed this using a known, challenging American finger alphabet dataset with about 60,000 depth images.

Handshape Recognition Using Skeletal Data.

In this paper, a method of handshapes recognition based on skeletal data is described. A new feature vector is proposed. It encodes the relative differences between vectors associated with the pointing directions of the particular fingers and the palm normal. Different classifiers are tested on the demanding dataset, containing 48 handshapes performed 500 times by five users. Two different sensor configurations and significant variation in the hand rotation are considered. The late fusion at the decision level of individual models, as well as a comparative study carried out on a publicly available dataset, are also included.