The paper proposes a visual tracking method based on deep learning without online model updating. In consideration of the advantages of deep learning in feature representation, deep model SSD (Single Shot Multibox Detector) is used as the object extractor in the tracking model. Simultaneously, the color histogram feature and HOG (Histogram of Oriented Gradient) feature are combined to select the tracking object. In the process of tracking, multi-scale object searching map is built to improve the detection performance of deep detection model and the tracking efficiency. In the experiment of eight respective tracking video sequences in the baseline dataset, compared with six state-of-the-art methods, the method in the paper has better robustness in the tracking challenging factors, such as deformation, scale variation, rotation variation, illumination variation, and background clutters, moreover, its general performance is better than other six tracking methods.
This paper describes a new method for infrared image segmentation based on mathematical morphology. The proposed
algorithm relies on four steps: First, to reduce the influence of asymmetrical background, top-hat transform was used,
and gradient image was obtained by morphological gradient transform. Second, gradient image was reconstructed by
reconstruction operator, which was constituted by doing opening by reconstruction operation and closing by
reconstruction operation successively. Through gradient image reconstruction, important region contours are preserved
while most tiny regular details and noises are removed. Finally, auto threshold technique was used to extract the region
edges form reconstructed gradient image. Experimental results show that the approach performs well in target
segmentation in infrared images with complicated background.