基于底层特征建模的行为识别算法优化

近年来基于时空兴趣点的视觉词袋(bag of video words,BOVW)模型被广泛用于行为识别算法研究;但是该模型忽略了每一种视觉单词的权重,另外没有考虑兴趣点时空分布信息,因而制约了其识别精度。提出了两种算法解决上述问题;其一,采用词频-逆向文件频率(term frequency-inverse document frequency,TF-IDF)方法对传统BOVW直方图进行优化处理,根据视觉单词在词袋与BOVW直方图的比例权衡其重要程度;其二,提出了基于三维共生矩阵的时空兴趣点互信息(spatialtemporal interest points mutual information,STIPsMI)算法,刻画不同视觉单词的时空兴趣点之间的时空关系。然后将STIPsMI描述符与优化后的BOVW直方图级联,作为视频序列最终的描述符。最后在两个主流的数据集KTH与UCF sports对该算法进行评估。实验结果表明,提出的时空特征描述符在行为识别准确率上优于BOVW模型与其他主流方法。     

Rapid detection of chromosome 18 aneuploidies in amniocytes by using primed in situ labeling （PRINS） technique

This paper presents a feasible method for rapid detection of the interphase nuclei of uncultured amniocytes for chromosomes 18 by using our modified in situ labeling (PRINS) technique.A total of 262 independent,uncultured amniotic fluid samples were analysed in a blind fashion before the karyotype was available.In addition,62 samples were examined by fluorescence in situ hybridization (FISH) for comparison.In more than 95% of the samples PRINS reactions with primer 18cen were successfully induced.Two samples were properly identified and correctly scored as trisomic 18.PRINS reaction could be performed automatically in less than one hour with a propgrammable thernocycler.Our studies showed that the PRINS technique is simple.rapid and cost-effective.It is as sensitive and specific as FISH;can enhance the eccuracy of standard cytogenetic analysis;and allows identification of chromosomes 18 aneuploidies in uncultrued amniocytes in significantly less time.