【摘要】：近年來,隨著機(jī)器視覺和人工智能技術(shù)的迅速發(fā)展,研究如何使機(jī)器具備空間場景理解和感知的能力,成為學(xué)者們研究的焦點(diǎn)。三維激光掃描測量技術(shù)的出現(xiàn)和發(fā)展為場景數(shù)據(jù)的獲取提供了全新的技術(shù)手段,它采用非接觸主動測量方式直接獲取場景高精度三維點(diǎn)云數(shù)據(jù),將機(jī)器對自然場景的理解和感知問題轉(zhuǎn)換成對點(diǎn)云的數(shù)據(jù)處理問題。但由于三維激光掃描測量技術(shù)獲取的點(diǎn)云數(shù)據(jù)具有海量、離散、缺乏語義信息等特性,因此,目前基于點(diǎn)云數(shù)據(jù)對場景的理解和感知仍是一項富有挑戰(zhàn)性的工作。依賴復(fù)雜精細(xì)的生物視覺系統(tǒng),人類天生具有對周圍環(huán)境感知和理解的能力,但機(jī)器要實現(xiàn)對場景的感知和理解,則要艱難的多,需要依賴各種算法和模型。基于場景點(diǎn)云的語義標(biāo)注方法,是當(dāng)前提高機(jī)器對場景理解和感知能力的熱點(diǎn)問題之一,也是本文的研究重點(diǎn)。該方法從整體上可以分為場景點(diǎn)云分割和場景點(diǎn)云標(biāo)注兩部分內(nèi)容,其中點(diǎn)云分割主要是將整體點(diǎn)云分割成互不重疊的點(diǎn)云區(qū)域,形成各自獨(dú)立的點(diǎn)云塊單元,從而實現(xiàn)對場景中物體的 感知‖,但是要達(dá)到對場景點(diǎn)云理解的目的,就需要對分割出來的點(diǎn)云塊進(jìn)行分類標(biāo)注。因此,本文的主要研究工作包括點(diǎn)云的分割和分類兩部分內(nèi)容,主要貢獻(xiàn)如下:(1)為了實現(xiàn)地面點(diǎn)云的分割,本文提出結(jié)合關(guān)鍵點(diǎn)的隨機(jī)采樣一致性地面點(diǎn)云分割算法。此算法在原有算法的基礎(chǔ)上增加多次迭代計算,統(tǒng)計獲得包含最多模型點(diǎn)的平面模型,并以這些點(diǎn)為關(guān)鍵點(diǎn)進(jìn)行平面擬合最終完成平面點(diǎn)云分割。改進(jìn)后的算法比原算法對于有起伏的地面點(diǎn)云有更好的分割效果,提高了場景點(diǎn)云中地面點(diǎn)云的分割效果。(2)為了實現(xiàn)場景點(diǎn)云中建筑立面的點(diǎn)云分割,本文使用改進(jìn)的區(qū)域增長算法來實現(xiàn)建筑立面點(diǎn)云分割。即在原有約束準(zhǔn)則歐式最小距離準(zhǔn)則的基礎(chǔ)上,增加種子點(diǎn)和鄰域點(diǎn)的法線角度這一判別條件,明顯改變了建筑立面點(diǎn)云數(shù)據(jù)與地面點(diǎn)云數(shù)據(jù)交接部位的分割結(jié)果,使得這兩部分的點(diǎn)云分割結(jié)果更加準(zhǔn)確。(3)為了更好和更完整的實現(xiàn)場景中樹木點(diǎn)云分割,本文在使用K-means聚類算法的基礎(chǔ)上結(jié)合圓柱體擬合算法實現(xiàn)了整顆樹木點(diǎn)云的分割。圓柱體擬合算法的加入,改變了只能分割樹冠部分點(diǎn)云的局面,使得對于樹木點(diǎn)云的分割更加完整和徹底。(4)借助分割獲得的獨(dú)立點(diǎn)云塊獲得高階團(tuán)構(gòu)建場景點(diǎn)云的條件隨機(jī)場模型,然后使用次梯度迭代算法和圖割推斷算法對條件隨機(jī)場模型的參數(shù)進(jìn)行學(xué)習(xí)和推斷,再結(jié)合場景點(diǎn)云的特征向量,最終實現(xiàn)場景點(diǎn)云的語義標(biāo)注。(5)在Visual Studio 2013開發(fā)平臺下,使用C++編程語言,對相應(yīng)的算法進(jìn)行編程實現(xiàn),最終達(dá)到對算法檢驗和驗證的目的。點(diǎn)云語義標(biāo)注方法的實現(xiàn),具有重大的現(xiàn)實意義,大大加快了數(shù)字化城市建模的速度并為機(jī)器自動導(dǎo)航提供了判別依據(jù),使機(jī)器獲得了對空間場景感知和理解的能力,具有廣闊的應(yīng)用前景和應(yīng)用價值。
[Abstract]:In recent years, with the rapid development of machine vision and artificial intelligence technology, the paper focuses on how to make the machine possess the ability of understanding and sensing space scene. the appearance and development of the three-dimensional laser scanning measurement technology provides a brand-new technical means for acquiring scene data, and adopts a non-contact active measurement mode to directly obtain the high-precision three-dimensional point cloud data of the scene, converting the understanding and perception of the machine to the natural scene into the data processing problem of the point cloud. However, it is still a challenging task to understand and perceive scene based on point cloud data because the cloud data acquired by three-dimensional laser scanning measurement technology has massive, discrete and lack of semantic information. Depending on the sophisticated biovision system, human beings are inherently capable of perception and understanding of the surrounding environment, but the machine needs to rely on algorithms and models to achieve the perception and understanding of the scene. Based on the semantic annotation method of the scene cloud, it is one of the hot issues to improve the understanding and perception ability of the machine at present, and it is also the focus of this paper. 璇ユ柟娉曚粠鏁翠綋涓婂彲浠ュ垎涓哄満鏅偣浜戝垎鍓插拰鍦烘櫙鐐逛簯鏍囨敞涓ら儴鍒嗗唴瀹，

本文編號：2270021