简介:Bycombininghistogramoforientedgradientandhistogramsofshearletcoefficients,whichanalyzesimagesatmultiplescalesandorientationsbasedonshearlettransforms,asthefeatureset,weproposedanovelhumandetectionfeature.Weemploypartialleastsquaresanalysis,anefficientdimensionalityreductiontechnique,toprojectthefeatureontoamuchlowerdimensionalsubspace.WetestitinINRIApersondatasetbyusingalinearSVM,andityieldsanerrorrateof1.38%withafalsenegatives(FN)rateof0.40%andafalsepositive(FP)rateof0.98%,whiletheerrorrateofHOGis7.11%,withaFNrateof4.09%andaFPrateof3.02%.
简介:通过分析ZigBee协议中Cluster—Tree和AODVjr算法的优缺点,提出一种基于Cluster—Tree+AODvjr的优化路由算法。该算法利用ZigBee协议中的邻居表,通过定义分区来确定目的节点的范围,从而控制广播RREQ分组的跳数,防止无效的RREQ泛洪。此优化算法能够有效地减小路由跳数,缩短传输时延,减少网络中死亡节点的数量,提高数据传送的成功率。