本文選題：線條痕跡 + 小波降噪　； 參考：《電子測量與儀器學(xué)報》2017年06期

【摘要】：高鐵線纜盜割案件案發(fā)現(xiàn)場遺留的斷頭承痕面上存在著大量線條痕跡,其往往呈現(xiàn)非線性形態(tài)特征,隨機(jī)性較強(qiáng)。為了更快速地進(jìn)行痕跡特征分析及所屬工具推斷,設(shè)計出一種針對非線性線條痕跡的小波域特征快速溯源算法,該算法首先將單點激光位移傳感器檢測斷頭表面拾取的一維信號利用小波分解進(jìn)行降噪,隨后使用基于小波特征的動態(tài)時間規(guī)整算法實現(xiàn)痕跡特征相似重合度匹配,最后使用基于梯度下降法的線性回歸機(jī)器學(xué)習(xí)算法不斷的迭代使得代價函數(shù)值代價最小,從而實現(xiàn)對應(yīng)工具快速推斷。最終通過實際痕跡推斷剪切工具試驗驗證了本算法的實用性和有效性。
[Abstract]:There are a large number of line marks on the face of the severed head mark left on the scene of the high speed wire and cable theft and cutting case, which often show nonlinear morphological characteristics and strong randomness. In order to analyze trace features more quickly and to infer their own tools, a fast traceability algorithm in wavelet domain for nonlinear line trace is designed. Firstly, the one-dimensional signal picked up by a single point laser displacement sensor is de-noised by wavelet decomposition, and then the dynamic time warping algorithm based on wavelet feature is used to match the similarity coincidence of trace features. Finally, the linear regression machine learning algorithm based on gradient descent method is used to iterate the cost function value to minimize the cost, so that the corresponding tool can be inferred quickly. Finally, the practicability and validity of the algorithm are verified by the actual trace inference and cutting tool test.
【作者單位】： 昆明理工大學(xué)航空學(xué)院;昆明理工大學(xué)機(jī)電工程學(xué)院;昆明信諾萊伯科技有限公司;
【基金】：云南省科技計劃(2014SC030,2016RA042,2017EH028) 公安部技術(shù)研究計劃(2014JSYJA020,2016JSYJA03) 昆明市科技計劃(2015-1-S-00284) 昆明理工大學(xué)分析測試基金(2016T20130030) 國家留學(xué)基金委創(chuàng)新型人才國際合作培養(yǎng)項目(201608740005)資助
【分類號】：TN911.7;U298
，

本文編號：2114999