基于Steiner点的移动目标遮挡恢复方法

Occluded moving object recovering based on Steiner point

Steiner点是物体的一个特征点,具有高稳定性和低离心性.由于其良好的鲁棒性,在跟踪预测移动目标位置有很好的效果,但Steiner点对移动物体边界有依赖,一旦移动物体部分被遮挡,Steiner点会丢失.为解决这一问题,从特征点的几何性质出发,通过推导有关遮挡的几何性质后,提出了一种解决遮挡问题的理论方法.首先提取移动物体的边界,经过凸壳处理、多边形逼近过程后,转化成易于处理的凸多边形.利用推导出的遮挡结论,计算出移动目标Steiner点的偏移向量,再结合未被遮挡区域的Steiner点,从而可以计算出移动目标的特征点——Steiner点的位置,从而解决被遮挡这种问题.通过一系列人工合成图像的实验,结果表明,目标移动部分被遮挡,Steiner点可以恢复出来.

The visibility of the moving target may be reduced by account that the target is shaded by the obstacle as it is moving.It will result in the loss of the critical information needed in our study and will decrease the reliability and accuracy of the tracking prediction.To overcome this problem,this paper proposes a new feature point method——moving Steiner point.Steiner point is a characteristic point of a object,with high stability and low eccentricity.Because of its robustness,tracking and predicting the location of moving target based on Steiner point have very good results.Once the object is partially occluded,then Steiner point will be lost.To recovery Steiner point,the definitions and related prosperities of the Steiner point are analyzed firstly,On the basis of its original mathematical definition,the concept of the moving Steiner point is presented and the continuity is proved.And the definition of the offset vector has been proposed.The mathematical relationship between the integral Steiner point and the topical Steiner point is established by taking advantage of the convex decomposion of Steiner point.By using the mathematical relationship,the occlusion model can be derived and the offset vector of the moving target＇s Steiner point can be calculated.Simultaneously,combined with the Steiner point of the unobstructed area,the characteristic value,namely the posi-tion of Steiner point,of the moving object can be deduced accordingly.The model has resolved the dilemma of numerous corners being obstructed during the movement of the target.In order to settle the problem that the boundary of the moving target is a complex curve,two algorithms have been proposed in the paper.Algorithm I,converting the complex curve into a polygon which is easier to handle.Algorithm II,studying the case of the unit circle being occluded,thus the occlusion problem can be converted into the problem of angle.A series of the synthetic images experiments shows that Steiner point can be restored when the moving object is occluded.