一种融合显著性的Adaboost的电视制导图像变尺度目标快速检测方法研究

电视制导过程中随着弹目距离的减小，目标图像会呈现尺度的非线性放大，现在传统的目标检测方法，不能满足目标尺度的精确变化，导致目标定位偏差，且检测计算量大，无法达到实时性。本文将基于剩余谱的显著性检测方法融入Adaboost检测过程，并应用于电视制导图像的目标检测中，能够对海面或沙漠等背景较为单一的区域进行快速目标检测，且检测窗口能够准确适应目标尺度的变化。

Research of TV Guided Image Fast Mutative Scale Object Detection Based On Adaboost Algorithm with Saliency Detection

With the decrease of the distance in TV guidance, target images would present the nonlinear bigger. It is usually use template scale update mechanism and gray correlation matching method for object detection. But the traditional method cannot Satisfy accurate scale change and lead to goal orientation deviation ,so it is not real-time because the amount of calculation is large. This paper apply Adaboost cascade classifier recognition method with spectral residual saliency detection to image in the target detection of TV guidance, the method the adoption of multiscale pyramid traversal, detection window to adapt to the changing of the target dimension accurately target detection.