Yujuan Qi

Memory-based Gaussian Mixture Modeling for moving object detection in indoor scene with sudden partial changes

In this paper, a memory-based Gaussian Mixture Model (MGMM) is proposed inspired by the way human perceives the environment. The human memory mechanism is introduced to model the background, which can make the model remember what the scene has ever been and help the model adapt to the variation of the scene more quickly. Experimental results show the effect of the memory mechanism in segmenting moving objects with sudden partial changes in the background scene.

Spinning tri-layer-circle memory modeling for template updating during moving object tracking

Inspired by the mechanism of human brain three-stage memory model, this paper develops a spinning tri-layer-circle memory model(STLC-MM) and applies it for template updating during object tracking. Three memory spaces are defined to store and process the object templates used in the tracking framework. Each memory space, which has a fixed input window and a fixed output window, is denoted by a circle and can spin with different speed. With three circle memory spaces spinning, templates in the three memory spaces are updated by imitating the cognitive process of memorization, recall, and forgetting. Then all the templates in the output windows of the three memory spaces are compared with the estimated template respectively, and the most similar template is selected as the final output of the STLC-MM. Finally, STLC-MM is incorporated into a particle filter (PF) framework in order to verify the effect of our proposed model. Experimental results show that the proposed method is more robust to sudden appearance changes and serious occlusions.

Biological modeling of human visual system for object recognition using GLoP filters and sparse coding on multi-manifolds

Hierarchical MAX model (HMAX) is a bio-inspired model mimicking the visual information processing of visual cortex. However, the visual processing of lower level, such as retina and lateral geniculate nucleus (LGN), is not concerned, and the properties of higher-level neurons are not sufficiently specified. Given that, we develop an extended HMAX model, denoted as E-HMAX, by the following biologically plausible ways. First, contrast normalization is conducted on the input image to simulate the processing of human retina and LGN. Second, log-polar Gabor (GLoP) filters are used to simulate the properties of V1 simple cells instead of Gabor filters. Then, sparse coding on multi-manifolds is modeled to compute the V4 simple cell response instead of Euclidean distance. Meanwhile, a template learning method based on dictionary learning on multi-manifolds is proposed to select informative templates during template learning stage. Experimental results demonstrate that the proposed model has greatly outperformed the standard HMAX model. It is also comparable to some state-of-the-art approaches such as EBIM and OGHM-HMAX.

Spinning tri-layer-circle memory-based Gaussian mixture model for background modeling

Inspired by the mechanism of human brain three-stage memory model and on the basis of our previous work, in this paper we present a novel spinning tri-layer-circle memory based Gaussian mixture model (STLCM-GMM). In this model, three circle memory spaces are defined to store and process the pixels and the Gaussians used in the segmentation framework respectively. With three circle memory spaces spinning, Gaussians in the three memory spaces are updated by imitating the cognitive process of memorization, recall, and forgetting. The proposed model could remember what the scene has ever been. When the similar scene occurs again, the model could adapt to the scene faster. The experimental results show the effectiveness of the proposed model in the field of background modeling.

Memory-based particle filter for real-time object tracking

Particle filter tracking algorithm based on global features becomes invalid when the targets appearance changes or is similar to the background. In order to solve such problems, we propose a memory-based particle filter which considers both local and global feature. Particles provide reliable matching area for local features so that error matching points can be eliminated. Then, local feature points matched to the target will guide the propagation of particles in order to avoid particle degeneration. Experimental results show the tracking effect of the proposed method under various conditions such as scale variation, sudden change of illumination, rotation and so on.

Memory-based codebook model for real-time object detection

In order to improve codebook (CB) model by reducing its training procedure, we introduce the human memory mechanism into CB and a memory-based codebook model is proposed. The proposed method extracts both the background and foreground into codewords and training procedure is no longer necessary. Experimental results show that our method improves the processing speed and achieves better performance in handling background variations. In addition, the proposed method can be applied in real-time monitoring.