基于感知哈希和自适应搜索的通用对象检测

Generic object detection based on perceptual hashing and selective search strate gy

摘要:

针对数字图像中对象检测问题,提出一种基于感知 哈希和搜索策略的通用 对象检测算法,提取图像中有可能包含对象的子区域并以矩形的形式输出。算法首先对输 入图像进行分割,并计算该图像的显著度信息;其次根据图像每个子区域显著度信息,自适 应 地确定搜索策略的起始子区域;最终利用选择性搜索策略,实现图像中通用对象检测。利用 选择性搜索策略检测通用对象时,引入感知哈希算法实现相邻子区域的相似度计算, 在保证准确度的基础上简化算法复杂度。实验结果显示,本文算法具有良好的图像对象检测 准确度。

Abstract:

To solve the problem of object detection,a generic object detection a lgorithm is proposed based on perceptual hashing and selective search strategy,a nd image regions that probably include the subdomain of objects are output in the form of rectangle.First of all,an input image is segmented,and the saliency information is computed.Secondl y,starting regions for selective search strategy is adaptively constructed based on saliency information and image segmentations.Finally,selective search strate gy is adopted to realize generic object detection within the input image.To improve the performance of selective search strategy,perceptual hashing method is introduced to compute similarity of each pair of neighbor regions,whi ch could reduce the complex of the computing process while maintaining the accuracy.The experimental results sh ow that the proposed algorithm has good accuracy of object detection.