Research on multi-view gait recognition using an omnidirectional camera

本研究试图提高步态识别的性能；从走路方式识别人。由于所获得的步态特征在不同的观察视图中是不同的，因此整合多视图步态特征可以提高步态识别的性能。以往的方法都是使用传统的透视相机来识别单视角的步态特征，而我们提出了一种使用全向相机的多视角步态识别方法。我们在室内的全向摄像机周围进行了各种观察视角步态图像序列的实验，验证了所提方法的有效性。此外，我们构建了步态识别性能模型，以实现识别性能最大化的最佳摄像机布局。

This study tried to improve performances of gait recognition ; person identification from walking style. Because obtained gait features are different among observation views, it is possible to improve the gait recognition performances integrating the multi-view gait features. While previous approaches use a single-view gait feature with a conventional perspective camera, we propose a method of multi-view gait recognition method by using an omnidirectional camera. We conducted experiments with various observation-view gait image sequences around the omnidirectional camera in the room and confirm the effectiveness of the proposed method. Moreover, we construct a gait recognition performance model to realize the optimal camera arrangement which maximize the recognition performance.