Application of Crowd Simulations in the Evaluation of Tracking Algorithms.

Tracking and action-recognition algorithms are currently widely used in video surveillance, monitoring urban activities and in many other areas. Their development highly relies on benchmarking scenarios, which enable reliable evaluations/improvements of their efficiencies. Presently, benchmarking methods for tracking and action-recognition algorithms rely on manual annotation of video databases, prone to human errors, limited in size and time-consuming. Here, using gained experiences, an alternative benchmarking solution is presented, which employs methods and tools obtained from the computer-game domain to create simulated video data with automatic annotations. Presented approach highly outperforms existing solutions in the size of the data and variety of annotations possible to create. With proposed system, a potential user can generate a sequence of random images involving different times of day, weather conditions, and scenes for use in tracking evaluation. In the design of the proposed tool, the concept of crowd simulation is used and developed. The system is validated by comparisons to existing methods.

Detection of linear features including bone and skin areas in ultrasound images of joints.

Identifying the separate parts in ultrasound images such as bone and skin plays a crucial role in the synovitis detection task. This paper presents a detector of bone and skin regions in the form of a classifier which is trained on a set of annotated images. Selected regions have labels: skin or bone or none. Feature vectors used by the classifier are assigned to image pixels as a result of passing the image through the bank of linear and nonlinear filters. The filters include Gaussian blurring filter, its first and second order derivatives, Laplacian as well as positive and negative threshold operations applied to the filtered images. We compared multiple supervised learning classifiers including Naive Bayes, k-Nearest Neighbour, Decision Trees, Random Forest, AdaBoost and Support Vector Machines (SVM) with various kernels, using four classification performance scores and computation time. The Random Forest classifier was selected for the final use, as it gives the best overall evaluation results.