摘要
针对胃上皮肿瘤细胞图像(以下简称肿瘤细胞图像)黏结严重和信息冗余的特点,提出了一种将自适应观测矩阵的压缩感知(SAM-CS)和自组织特征映射(SOFM)神经网络相结合的算法。该算法将肿瘤细胞图像拉成列向量,然后利用通过自适应过程产生的观测矩阵,基于压缩感知理论对图像信息进行观测,产生线性观测向量,最后利用SOFM神经网络的学习算法对观测向量进行训练和分类,实现对肿瘤细胞图像的识别。实验表明,相比常用算法,该算法至少提高了4.2%的识别准确率和5.7%的运算速度。
Given the characteristics of serious cementation and information redundancy of gastric epithelium tumor cell images (hereinafter referred to as tumor cell images), we propose an algorithm which is a combination of the compressed sensing of self-adaptive measurement (SAM-CS) matrix and the selforganizing feature map (SOFM) neural network. Firstly, the tumor cell images are transferred to column vectors, then the linear observation vectors are generated through the SAM-CS theory. Finally, we train and classify the linear observation vectors by using the learning algorithm of SOFM neural network to implement the recognition of tumor cell images. Experimental results show that compared with traditional algorithms, the proposed algorithm has improved 4.2 % of the recognition accuracy and 5.7% of the operation speed at least.
出处
《计算机工程与科学》
CSCD
北大核心
2015年第8期1558-1565,共8页
Computer Engineering & Science
基金
国家自然科学基金资助项目(61163040
61402227)
江西省教育厅资助项目(GJJ10451
GJJ14372)
关键词
自适应观测矩阵
压缩感知
自组织特征映射
肿瘤细胞图像识别
self-adaptive measurement matrix
compressed sensing
self-organizing feature map
the recognition of tumor cell images