Probabilistic Appearance Model Research Articles

Enhanced anomaly detection in wire ropes by combining structure and appearance

Automatic visual surface inspection is a challenging task, which has become important for quality assurance in the last years. Wire rope inspection is a special problem within this field. Usually, the huge and heavy ropes cannot be detached. Thus, an inspection of the ropes must be conducted, while the ropes are in use. The rope surface exhibits various appearance characteristics so that the existing, purely appearance-based approaches tend to fail.We explicitly integrate information about the object geometry, which we obtain by aligning a sequence of 2d rope images with a perfectly regular 3d model of the rope. The rendering equation is used to link object geometry to the observed rope appearance. Based on the connection between geometry and surface appearance we build a probabilistic appearance model which serves as representation for normal surface variations. A robust localization of rope surface defects is achieved by means of anomaly detection. The presented approach has no need for knowledge about the illumination setting or the reflectance properties of the material.An evaluation on real-world data from ropeways leads to an accuracy comparable to that of a human expert. With an accuracy of 95% and a false-alarm-rate of 1.5% the approach outperforms all other existing approaches.

Human Detection from Fish-eye Image Based on Probabilistic Appearance Model

Human Detection from Fish-eye Image Based on Probabilistic Appearance Model

Multi-target tracking with occlusions via skeleton points assignment

Multiple-target tracking in complex scenes is one of the most complicated problems in computer vision. Handling the occlusion between objects is the key issue in multiple-target tracking. This paper introduces the method of motion segmentation into the object tracking system, and presents a SPA (Skeleton Points Assign, SPA) based occlusion segmentation approach to track multiple people through complex situations captured by static monocular cameras. In the proposed method, we first select the skeleton points and evaluate their occlusion states by low-level information like optical flow; then we assign these points to different objects using advanced semantic information, such as appearance, motion and color; finally, a dense classification of foreground pixels are taken advantages of to accomplish occlusion segmentation and a blob-based compensation strategy is utilized to estimate the missing information of occluded objects. Object tracking is handled by a particle filter-based tracking framework, in which a probabilistic appearance model is used to find the best particle. Experiments are conducted on the public challenging dataset PETS 2009. Results show that this approach can improve the performance of the existing tracking approach and handle dynamic occlusions better.

A low complexity approximation of probabilistic appearance models

Appearance models yield a compact representation of shape, pose and illumination variations. The probabilistic appearance model, introduced by Moghaddam et al. (Proceedings of the International Conference on Computer Vision, Cambridge, MA, June 1995, p. 687; IEEE Trans. Pattern Anal. Mach. Intell. 19 (7) (1997) 696) has recently shown excellent performances in pattern detection and recognition, outperforming most other linear and non-linear approaches. Unfortunately, the complexity of this model remains high. In this paper, we introduce an efficient approximation of this model, which enables fast implementations in statistical estimation-based schemes. Gains in complexity and cpu time of more than 10 have been obtained, without any loss in the quality of the results.