Research on Target Detection of Aerial Photography Based on Combination Image Signature with Frequency-Tuned Saliency Feature

To improve efficiency and reliability of target detection of aerial photography under uncertainty such as atmosphere and light, we propose a novel detection method based on combination image signature with frequency-tuned saliency feature. Firstly, a descriptor known as image signature is used to extract target regions by discrete cosine transformation in frequency domain. Secondly, frequency-tuned saliency highlights the important edges and content of target area. Then, Riesz descriptor is used to extract region edge and suppress the noise, and prior Hough transformation is used to extract the boundary line features. Finally, the target is got on reliance and high efficiency. The experimental results demonstrate that the proposed method is effective and efficiency.