基于仿生视觉注意机制的无人机目标检测

Unmanned Aerial Vehicles Target Detection Based on Bio-inspired Visual Attention

提出了一种基于仿生视觉注意机制的无人机目标检测方法,该方法使用了亮度、方向和区域对比度特征,针对提取的多个显著性特征,利用Ada Boost分类器对其进行分析和融合,得到最终的显著图。对显著图进行图像分割,从中找出显著度最高的区域即目标区域。仿真结果表明,所提出的无人机目标检测方法可以比较准确地确定目标区域,自适应能力强。

A novel method of target detection for unmanned aerial vehicles （UAV） was proposed based on the mechanism of biological visual attention. Different features including intensity, orientation and region contract were utilized in the proposed algorithm. The AdaBoost clasififer was utilized to analyze the saliency features and fuse the feature maps into the ifnal saliency map. Then, target region that has the largest value in the saliency map was detected by image segmentation. Series of experimental results demonstrate the feasibility and effectiveness of the proposed approach of target detection for UAVs, which has high adaptive ability and can detect the target region precisely.