本文選題：場景化 + 人臉識別��； 參考：《北京交通大學(xué)》2017年碩士論文

【摘要】：眼下,正是一個智能機器人爆發(fā)的時代,無論是工業(yè)機器人還是特種機器人、家用機器人,機器人時代到來正帶給現(xiàn)代社會和現(xiàn)代人各種驚喜。對于現(xiàn)代社會中忙碌于工作的年輕父母們而言,他們利用機器人為孩子提供更好陪伴的夢想也正在變得觸手可及�；跈C器人場景交互的人臉識別系統(tǒng)是兒童教育機器人的一個重要模塊,目的是讓機器人識別不同的人物角色,從而有目的進行交互,使得機器人更加智能化,更容易融入到人們的生活場景中�；跈C器人場景交互的人臉識別系統(tǒng)是應(yīng)用在人們生活的具體場景之中。生活場景是豐富多樣的,所以要想讓機器人自然的融入人類的生活場景中,需要盡可能的豐富其場景處理能力。這個問題需要從兩個方面來解決,一方面可以盡量多的增加場景邏輯,考慮到生活中盡可能多的場景;另一方面,提高場景的適用性,用少量的場景邏輯來處理現(xiàn)實中多個場景。本文使用人臉識別相關(guān)技術(shù),來實現(xiàn)機器人場景化中的認(rèn)識新朋友和與好朋友打招呼兩個場景。首先,本文介紹了系統(tǒng)的實際應(yīng)用背景和人臉識別在系統(tǒng)中的應(yīng)用。同時,也介紹了人臉識別的相關(guān)背景和國內(nèi)外的發(fā)展現(xiàn)狀。其次,結(jié)合系統(tǒng)的業(yè)務(wù)需求,確定了系統(tǒng)的總體架構(gòu)和模塊劃分,以及各個模塊之間的交互流程。最后,本文詳細闡述了基于機器人場景交互的人臉識別系統(tǒng)的實現(xiàn)過程,該部分是本文的重點。本文的創(chuàng)新點包括以下幾個方面:首先,本文將人臉識別算法應(yīng)用在機器人與人類實際交互的場景中,具有實際的應(yīng)用意義;其次,在特征提取的算法中,將CNN結(jié)構(gòu)的倒數(shù)第二層全連接層的結(jié)果作為人臉的特征保存,這種方法不僅保留了 CNN的識別準(zhǔn)確度,又提高了特征提取的效率;最后,將場景化實現(xiàn)與視覺算法實現(xiàn)分離,提高了系統(tǒng)的可擴展性。
[Abstract]:At present, it is an era of intelligent robot explosion, whether industrial robot or special robot, home robot, robot era is bringing modern society and modern people a variety of surprises. For the busy young parents of modern society, their dream of using robots to provide better companionship for their children is also becoming within reach. The human face recognition system based on robot scene interaction is an important module of children education robot. The purpose of this system is to let the robot recognize different personas, so that the robot can interact purposefully and make the robot more intelligent. It's easier to fit into people's life scenes. The human face recognition system based on robot scene interaction is applied in the concrete scene of people's life. The life scene is rich and diverse, so if we want the robot to integrate into the human life scene naturally, we need to enrich its scene processing ability as much as possible. This problem needs to be solved from two aspects. On the one hand, the logic of the scene can be increased as much as possible, taking into account as many scenes as possible in life; on the other hand, the applicability of the scene can be improved. Use a small amount of scenario logic to handle multiple scenarios in reality. In this paper, face recognition related techniques are used to realize the two scenes of new friends and good friends in robot scene. Firstly, this paper introduces the practical application background of the system and the application of face recognition in the system. At the same time, it also introduces the background of face recognition and the development situation at home and abroad. Secondly, according to the business requirements of the system, the overall architecture and module partition of the system are determined, as well as the interaction flow between each module. Finally, this paper describes the realization process of the robot scene interactive face recognition system in detail, which is the focus of this paper. The innovation of this paper includes the following aspects: first, this paper applies face recognition algorithm to the actual interaction between robot and human, which has practical application significance; secondly, in the feature extraction algorithm, This method not only preserves the recognition accuracy of CNN, but also improves the efficiency of feature extraction. Finally, the realization of scene is separated from the realization of visual algorithm. The expansibility of the system is improved.
【學(xué)位授予單位】：北京交通大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP391.41;TP242

【參考文獻】

相關(guān)期刊論文 前10條

1 嚴(yán)嚴(yán);陳日偉;王菡子;;基于深度學(xué)習(xí)的人臉分析研究進展[J];廈門大學(xué)學(xué)報(自然科學(xué)版);2017年01期

2 鄒修明;孫懷江;楊賽;;基于概率隱含語義分析模型的人臉識別算法[J];南京理工大學(xué)學(xué)報;2016年05期

3 邢玲;馮倩;穆國旺;;基于卷積神經(jīng)網(wǎng)絡(luò)和PCA的人臉識別[J];河北工業(yè)大學(xué)學(xué)報;2016年05期

4 余丹;吳小俊;;一種卷積神經(jīng)網(wǎng)絡(luò)和極限學(xué)習(xí)機相結(jié)合的人臉識別方法[J];數(shù)據(jù)采集與處理;2016年05期

5 劉萬軍;梁雪劍;曲海成;;基于雙重優(yōu)化的卷積神經(jīng)網(wǎng)絡(luò)圖像識別算法[J];模式識別與人工智能;2016年09期

6 陳耀丹;王連明;;基于卷積神經(jīng)網(wǎng)絡(luò)的人臉識別方法[J];東北師大學(xué)報(自然科學(xué)版);2016年02期

7 朱陶;任海軍;洪衛(wèi)軍;;一種基于前向無監(jiān)督卷積神經(jīng)網(wǎng)絡(luò)的人臉表示學(xué)習(xí)方法[J];計算機科學(xué);2016年06期

8 文孟飛;胡超;劉偉榮;;一種基于深度學(xué)習(xí)的異構(gòu)多模態(tài)目標(biāo)識別方法[J];中南大學(xué)學(xué)報(自然科學(xué)版);2016年05期

9 胡亞君;;卷積神經(jīng)網(wǎng)絡(luò)在人臉識別的應(yīng)用[J];無線互聯(lián)科技;2016年09期

10 張丹丹;李雷;;基于PCANet-RF的人臉檢測系統(tǒng)[J];計算機技術(shù)與發(fā)展;2016年02期

，

本文編號：1945806