基于稀疏超完备表示的目标检测算法

Object detection algorithm based on sparse overcomplete representation

基于视觉超完备机制的图像稀疏表示是一种新的图像表征方法。针对目标检测问题,提出了一种基于视觉稀疏超完备表示的计算模型,实现了非结构化场景中的目标检测。该方法首先基于能量模型和评分匹配(score matching)方法建立稀疏超完备计算模型,进而设计了基于神经元响应以及动态阈值策略的目标检测算法,最后通过多类型交通图像验证算法有效性。结果表明,该方法与计算机视觉方法比较具有较高的准确率,能够利用少样本实现大交通流量中目标的检测。

Image sparse representation based on visual overcomplete mechanism is a new image representation meth- od. Aiming at the object detection problem in image, we propose a computing model based on visual sparse overcom- plete representation method, which achieves the object detection in unstructured scenes. In this method, firstly, a sparse overcomplete computing model is established based on energy model and score matching method. Then the ob- ject detection algorithm based on neuron response and dynamic threshold strategies is designed. Various kinds of traf- fic images were used to verify the effectiveness of the algorithm. The results show that compared with computer vision methods, the proposed method has higher accuracy and can realize the object detection in large traffic flow with fewer samples.