摘要
由于图像中的背景复杂、目标弱小,在面对具有强噪声或者目标存在遮掩的情况下没有很好的检测算法。提出一种改进的Laplacian算子和Hog特征及Adaboost分类器相结合的算法。该方法根据拉普拉斯算子的边缘检测特性,先通过中值滤波滤掉一部分随机噪声,然后利用改进的Laplacian算子对图像进行卷积,获取图像使用稀疏矩阵表示的Hog特征,输入到Adaboost分类器后最终得到一个最准确的目标,并且在仿真实验中的准确率高达93.5%。
Due to complex background in the image and the weak target,there is no good detection algorithm in the presence of strong noise or concealment of the target.The article proposes an improved algorithm combining Laplacian operator with Hog feature and Adaboost classifier.According to the edge detection characteristics of the Laplacian operator,this method firstly filters out a part of random noise through median filtering,and then uses the improved Laplacian operator to convolve the image to obtain the Hog feature represented by the sparse matrix of the image,and input it to the Adaboost classifier for obtaining the most accurate target,and the accuracy rate in the simulation experiment is as high as 93.5%.
作者
刘镏
匡迎春
沈岳
周浩宇
LIU Liu;KUANG Yingchun;SHEN Yue;ZHOU Haoyu(College of Informatioii Science and Technology,Hunan Agricultural University,Changsha 410128,China)
出处
《激光杂志》
CAS
北大核心
2022年第5期65-69,共5页
Laser Journal
基金
国家自然科学基金(No.61972147)
国家科技计划支撑项目(No.2012BAD35B05)。
