Security and safety applications such as surveillance or forensics demand face recognition in low-resolution video data. We propose a face recognition method based on a Convolutional Neural Network (CNN) with a manifold-based track comparison strategy for low-resolution video face recognition. The low-resolution domain is addressed by adjusting the network architecture to prevent bottlenecks or significant upscaling of face images. The CNN is trained with a combination of a large-scale self-collected video face dataset and large-scale public image face datasets resulting in about 1.4M training images. To handle large amounts of video data and for effective comparison, the CNN face descriptors are compared efficiently on track level by local patch means. Our setup achieves 80.3 percent accuracy on a 32×32 pixels low-resolution version of the YouTube Faces Database and outperforms local image descriptors as well as the state-of-the-art VGG-Face network [20] in this domain. The superior performance of the proposed method is confirmed on a self-collected in-the-wild surveillance dataset.

Low-resolution face analysis suffers more significantly from quality degradations than high-resolution analysis. In this work, we will investigate how several face analysis steps are influenced by low image quality and how this relates to the low resolution. In the first step, a simulation of different effects on image quality, namely low resolution, compression artifacts, motion blur and noise is performed and the impact on face detection, registration and recognition is analyzed. Depending on the situation, it becomes obvious that the low resolution is sometimes a minor degrading effect, outmatched by a single one or a combination of the further effects. When addressing real-world face recognition from surveillance data, the combination of the challenging effects is the biggest problem because typical counter measures are individual to one single effect.

This paper establishes a duality between the calculus of variations, an increasingly common method for trajectory planning, and Hidden Markov Models (HMMs), a common probabilistic graphical model with applications in artificial intelligence and machine learning. This duality allows findings from each field to be applied to the other, namely providing an efficient and robust global optimization tool and machine learning algorithms for variational problems, and fast local solution methods for large state-space HMMs.

Performance of perceptive systems depends on the quality of the input data. In this contribution, an approach to evaluate perception performance as a function of quality of the sensor data is presented. Standardized quality metrics support the imaging sensors performance measurement. Several imaging setups are analyzed with real world experiments. The output of each setup is processed offline to track down performance differences with respect to the quality of sensor data. An adapted measurement is calculated to measure the sensor performance with respect to the data quality for the involved perceptive components. The measured performance is assessed by processing the data of different simultaneously recorded imaging setups for the task of feature extraction of road lanes.

A new algorithm for building irregular pyramids is presented. The algorithm is based on only two basic operations on graphs, contraction and removal of edges. By making use of the concept of dual graphs, the algorithm overcomes the problem of unbounded vertex degree proper to the existing approach to building irregular pyramids. This boundedness extends the scope of parallel, degree preserving graph contraction also to irregular structures. Our method can be applied to all problems in which a 2D discrete space can be represented by a planar graph. Four-connected square grids, region adjacency graphs, Voronoi diagrams and Delaunay triangulations are examples of such plane representations.

In order to usefully apply the concept of the curve pyramid on grouping problems, the shift variance problem must be overcome by extending the concept to irregular pyramids. These have a structure that adapts to the image data by deriving control information from curve relations. The algorithm that builds the irregular curve pyramid by deriving higher levels of abstraction from a set of relations goes far beyond merely solving the shift variance problem. It can reduce the computational complexity in comparable applications where all possible combinations of parts have to be checked in order to reassemble complex objects.>