Ulsee Inc.

Jingjing Xiao

An image target tracking method and system thereof are provided in the present disclosure. The image target tracking method includes the following steps: obtaining a target initial position, and performing a sparse sampling according to the target initial position; dividing sampling points into foreground sampling points and background sampling points; clustering adjacent foreground sampling points according to a spatial distribution of the foreground sampling points in order to obtain a clustering result containing a plurality of clusters; performing a robust estimation according to the clustering result in order to determine a relative position between a target and a camouflage interference in an image; and generating a prediction trajectory, correlating an observation sample position with the prediction trajectory to generate a correlation result, and determining whether the target is blocked and tracking the target according to the correlation result.

Image target tracking method and system thereof

We propose a tracking framework that explicitly encodes both generic features and category-based features. The tracker consists of a shared convolutional network (NetS), which feeds into two parallel networks, NetC for classification and NetT for tracking. NetS is pre-trained on ImageNet to serve as a generic feature extractor across the different object categories for NetC and NetT. NetC utilizes those features within fully connected layers to classify the object category. NetT has multiple branches, corresponding to multiple categories, to distinguish the tracked object from the background. Since each branch in NetT is trained by the videos of a specific category or groups of similar categories, NetT encodes category-based features for tracking. During online tracking, NetC and NetT jointly determine the target regions with the right category and foreground labels for target estimation.

Target tracking with inter-supervised convolutional networks

An image target tracking method, device, and system thereof are provided in the present disclosure. The image target relative position determining method includes the following steps: obtaining a target initial position, and performing a sparse sampling according to the target initial position; dividing sampling points into foreground sampling points and background sampling points; clustering adjacent foreground sampling points according to a spatial distribution of the foreground sampling points in order to obtain a clustering result containing a plurality of clusters; and performing a robust estimation according to the clustering result in order to determine a relative position between a target and a camouflage interference in an image. Throughout the process, a multi-feature cascade clustering is completed by using sparse sampling, sampling point division, and adjacent foreground sampling point clustering; a robust estimation is performed in order to accurately predict a relative position between a target and a camouflage interference.

Image target relative position determining method, device, and system thereof

An image target tracking method and system thereof are provided in the present disclosure. The image target tracking method includes the following steps: determining a relative position between a target and a camouflage interference in an image; generating a prediction trajectory according to the relative position between the target and the camouflage interference in the image; and correlating an observation sample position with the prediction trajectory to generate a correlation result, and determining whether the target is blocked and tracking the target according to the correlation result. Throughout the process, the prediction trajectory is generated based on the determined relative position between the target and the camouflage interference, and the prediction trajectory is correlated to determine whether the target is blocked and to accurately track the target.

Growing community of inventors

Beijing, China

Jingjing Xiao