Semantic feature-based visual attention model for pedestrian detection

Objective
Pedestrian detection under video surveillance systems has always been a hot topic in computer vision research. These systems are widely used in train stations, airports, large commercial plazas, and other public places. However, pedestrian detection remains difficult because of complex backgrounds. Given its development in recent years, the visual attention mechanism has attracted increasing attention in object detection and tracking research, and previous studies have achieved substantial progress and breakthroughs. We propose a novel pedestrian detection method based on the semantic features under the visual attention mechanism.
Method
The proposed semantic feature-based visual attention model is a spatial-temporal model that consists of two parts: the static visual attention model and the motion visual attention model. The static visual attention model in the spatial domain is constructed by combining bottom-up with top-down attention guidance. Based on the characteristics of pedestrians, the bottom-up visual attention model of Itti is improved by intensifying the orientation vectors of elementary visual features to make the visual saliency map suitable for pedestrian detection. In terms of pedestrian attributes, skin color is selected as a semantic feature for pedestrian detection. The regional and Gaussian models are adopted to construct the skin color model. Skin feature-based visual attention guidance is then proposed to complete the top-down process. The bottom-up and top-down visual attentions are linearly combined using the proper weights obtained from experiments to construct the static visual attention model in the spatial domain. The spatial-temporal visual attention model is then constructed via the motion features in the temporal domain. Based on the static visual attention model in the spatial domain, the frame difference method is combined with optical flowing to detect motion vectors. Filtering is applied to process the field of motion vectors. The saliency of motion vectors can be evaluated via motion entropy to make the selected motion feature more suitable for the spatial-temporal visual attention model.
Result
Standard datasets and practical videos are selected for the experiments. The experiments are performed on a MATLAB R2012a platform. The experimental results show that our spatial-temporal visual attention model demonstrates favorable robustness under various scenes, including indoor train station surveillance videos and outdoor scenes with swaying leaves. Our proposed model outperforms the visual attention model of Itti, the graph-based visual saliency model, the phase spectrum of quaternion Fourier transform model, and the motion channel model of Liu in terms of pedestrian detection. The proposed model achieves a 93% accuracy rate on the test video.
Conclusion
This paper proposes a novel pedestrian method based on the visual attention mechanism. A spatial-temporal visual attention model that uses low-level and semantic features is proposed to calculate the saliency map. Based on this model, the pedestrian targets can be detected through focus of attention shifts. The experimental results verify the effectiveness of the proposed attention model for detecting pedestrians.

Video anomaly detection in real-time on a Power-Aware Heterogeneous Platform

FPGAs and GPUs are often used when real-time performance in video processing is required. An accelerated processor is chosen based on task-specific priorities (power consumption, processing time and detection accuracy), and this decision is normally made once at design time. All three characteristics are important, particularly in battery-powered systems. Here we propose a method for moving selection of processing platform from a single design-time choice to a continuous run time one.We implement Histogram of Oriented Gradients (HOG) detectors for cars and people and Mixture of Gaussians (MoG) motion detectors running across FPGA, GPU and CPU in a heterogeneous system. We use this to detect illegally parked vehicles in urban scenes. Power, time and accuracy information for each detector is characterised. An anomaly measure is assigned to each detected object based on its trajectory and location, when compared to learned contextual movement patterns. This drives processor and implementation selection, so that scenes with high behavioural anomalies are processed with faster but more power hungry implementations, but routine or static time periods are processed with power-optimised, less accurate, slower versions. Real-time performance is evaluated on video datasets including i-LIDS. Compared to power-optimised static selection, automatic dynamic implementation mapping is 10% more accurate but draws 12W extra power in our testbed desktop system.

Experimental evaluation of the response of micro-channel plate detector to ions with 10s of MeV energies

The absolute calibration of a microchannel plate (MCP) assembly using a Thomson spectrometer for laser-driven ion beams is described. In order to obtain the response of the whole detection system to the particles’ impact, a slotted solid state nuclear track detector (CR-39) was installed in front of the MCP to record the ions simultaneously on both detectors. The response of the MCP (counts/particles) was measured for 5–58 MeV carbon ions and for protons in the energy range2–17.3 MeV. The response of the MCP detector is non-trivial when the stopping range of particles becomes larger than the thickness of the detector. Protons with energiesE>~ 10 MeV are energetic enough that they can pass through the MCP detector. Quantitative analysis of the pits formed in CR-39 and the signal generated in the MCP allowed to determine the MCP response to particles in this energy range. Moreover, a theoretical model allows to predict the response of MCP at even higher proton energies. This suggests that in this regime the MCP response is a slowly decreasing function of energy, consistently with the decrease of the deposited energy. These calibration data will enable particle spectra to be obtained in absolute terms over a broad energy range.

Spectral analysis of four 'hypervariable' AGN: a micro-needle in the haystack?

We analyze four extreme AGN transients to explore the possibility that they are caused by rare, high-amplitude microlensing events. These previously unknown type-I AGN are located in the redshift range 0.6-1.1 and show changes of > 1.5 magnitudes in the g-band on a timescale of ~years. Multi-epoch optical spectroscopy, from the William Herschel Telescope, shows clear differential variability in the broad line fluxes with respect to the continuum changes and also evolution in the line profiles. In two cases a simple point-source, point-lens microlensing model provides an excellent match to the long-term variability seen in these objects. For both models the parameter constraints are consistent with the microlensing being due to an intervening stellar mass object but as yet there is no confirmation of the presence of an intervening galaxy. The models predict a peak amplification of 10.3/13.5 and an Einstein timescale of 7.5/10.8 years respectively. In one case the data also allow constraints on the size of the CIII] emitting region, with some simplifying assumptions, to to be ~1.0-6.5 light-days and a lower limit on the size of the MgII emitting region to be > 9 light-days (half-light radii). This CIII] radius is perhaps surprisingly small. In the remaining two objects there is spectroscopic evidence for an intervening absorber but the extra structure seen in the lightcurves requires a more complex lensing scenario to adequately explain.