【摘要】：物體檢測是計算機(jī)視覺研究領(lǐng)域中一個極富挑戰(zhàn)性的課題,是大量高級視覺任務(wù)的基礎(chǔ)。盡管經(jīng)歷了數(shù)十年的研究發(fā)展,但在面對實際復(fù)雜變化場景時,物體檢測的表現(xiàn)仍然存在不足。作為兼具分類和定位任務(wù)的復(fù)雜系統(tǒng)問題,物體檢測始終在模型能力和計算代價的取舍中徘徊前進(jìn)。針對一般物體檢測,本文依據(jù)識別對象的幾何變化,將目標(biāo)分為結(jié)構(gòu)化和非結(jié)構(gòu)化目標(biāo)。結(jié)構(gòu)化物體檢測的核心問題是如何表達(dá)目標(biāo)的幾何結(jié)構(gòu)信息,如何建模結(jié)構(gòu)化物體的幾何變化。針對結(jié)構(gòu)化物體檢測,本文假設(shè)目標(biāo)對象的幾何變化是透視變換,使用點特征集合表達(dá)目標(biāo)的幾何結(jié)構(gòu),使用S-SVM分類器建模結(jié)構(gòu)化物體檢測算法。本文提出預(yù)訓(xùn)練和跟蹤算法進(jìn)一步提高結(jié)構(gòu)化物體檢測效率。實驗結(jié)果表明預(yù)訓(xùn)練能提高分類器對同一類點特征的辨別力,跟蹤算法在不嚴(yán)重?fù)p失準(zhǔn)確度的情況下能大幅提高檢測速度。非結(jié)構(gòu)化物體檢測的核心問題是如何表達(dá)目標(biāo)區(qū)域信息,如何將候選區(qū)域的提取與物體分類統(tǒng)一建模優(yōu)化。針對非結(jié)構(gòu)化物體檢測,本文使用基于特征學(xué)習(xí)的數(shù)據(jù)驅(qū)動特征來表達(dá)目標(biāo)區(qū)域信息,使用Faster R-CNN建模非結(jié)構(gòu)化物體檢測算法。本文提出基于多層刺激的候選框融合算法進(jìn)一步提高非結(jié)構(gòu)化物體檢測效率。實驗結(jié)果表明多層刺激算法能進(jìn)一步豐富特征抽象能力,該特征使得學(xué)習(xí)器學(xué)習(xí)出更加魯棒的分類規(guī)則。綜上所述,本文分析了結(jié)構(gòu)化與非結(jié)構(gòu)化的物體檢測方法,并提出相應(yīng)的改進(jìn)算法,提高了在特定應(yīng)用場景下的物體檢測效率。
[Abstract]:Object detection is a very challenging task in the field of computer vision. It is the basis of a large number of advanced visual tasks. Despite decades of research and development, the performance of object detection still exists in the face of actual complex change scenes. As a complex system problem with both classification and location tasks, physical examination is a problem. For general object detection, this paper divides the object into structured and unstructured object according to the geometric change of the object. The core problem of structured object detection is how to express the geometric structure of the target and how to model the geometric change of the structured object. In view of structured object detection, this paper assumes that the geometric change of the target object is perspective transformation, uses the point feature set to express the geometric structure of the target, and uses the S-SVM classifier to model the structured object detection algorithm. This paper proposes a pre training and tracking algorithm to further improve the physical examination efficiency of the structuration. The experimental results show that the pre training can be proposed. The recognition ability of the high classifier to the same kind of point features, the tracking algorithm can greatly improve the detection speed when the accuracy of the loss is not serious. The core problem of the unstructured object detection is how to express the target area information, how to model the extraction of the candidate region and the object classification, and to detect the unstructured objects. The data driven feature based on feature learning is used to express the target area information, and the Faster R-CNN is used to model the unstructured object detection algorithm. In this paper, a candidate frame fusion algorithm based on multi-layer stimulation is proposed to further improve the detection efficiency of unstructured objects. The experimental results show that the multi-layer stimulation algorithm can further enrich the feature abstraction energy. This feature makes the learner learn more robust classification rules. In summary, this paper analyzes structural and unstructured object detection methods, and proposes a corresponding improvement algorithm to improve the efficiency of object detection in a specific application scene.
【學(xué)位授予單位】：浙江大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP391.41

【相似文獻(xiàn)】

相關(guān)期刊論文 前10條

1 鮑培明;模糊Petri網(wǎng)在非結(jié)構(gòu)化決策支持中的應(yīng)用研究[J];計算機(jī)工程;2001年12期

2 劉晴;;昆騰創(chuàng)新歸檔解決方案減輕了主存儲上非結(jié)構(gòu)化數(shù)據(jù)增長的負(fù)擔(dān)[J];計算機(jī)與網(wǎng)絡(luò);2013年23期

3 繆紅;鐘華;龍昕;;元器件企業(yè)保護(hù)非結(jié)構(gòu)化大數(shù)據(jù)的研究[J];電子元件與材料;2014年06期

4 王曉麗;;基于網(wǎng)絡(luò)的中學(xué)非結(jié)構(gòu)化教育資源建設(shè)探析[J];中國教育信息化;2007年08期

5 ToddMatsler;;深耕大數(shù)據(jù) 助力平安城市智慧轉(zhuǎn)型[J];中國信息界;2013年12期

6 經(jīng)有國;但斌;張旭梅;郭鋼;;基于本體的非結(jié)構(gòu)化客戶需求智能解析方法[J];計算機(jī)集成制造系統(tǒng);2010年05期

7 張廣泉;非結(jié)構(gòu)化程序流程圖及其等價變換[J];重慶師范學(xué)院學(xué)報(自然科學(xué)版);1993年03期

8 王曉波;;非結(jié)構(gòu)化數(shù)據(jù)采集和檢索技術(shù)的研究與實現(xiàn)[J];中國內(nèi)部審計;2014年07期

9 黃遠(yuǎn)鳴;;一種非結(jié)構(gòu)化數(shù)據(jù)的多牽度分配存取實現(xiàn)方法[J];科技通報;2014年08期

10 劉威,武家春,廖建新,丁軼;非結(jié)構(gòu)化補充業(yè)務(wù)數(shù)據(jù)中心的設(shè)計與實現(xiàn)[J];計算機(jī)工程與應(yīng)用;2005年10期

相關(guān)會議論文 前10條

1 祝世京;陳s，

本文編號：2167206