本文關(guān)鍵詞：基于OpenCV運(yùn)動目標(biāo)檢測與跟蹤方法研究　出處：《沈陽航空航天大學(xué)》2017年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 目標(biāo)檢測 目標(biāo)跟蹤 背景建模 粒子濾波 Kalman算法 MeanShift

【摘要】：視頻中運(yùn)動目標(biāo)檢測和跟蹤是計(jì)算機(jī)視覺和模式識別領(lǐng)域的研究熱點(diǎn),在智能視頻監(jiān)控系統(tǒng)、人工智能、視覺導(dǎo)航等方面有著廣泛的應(yīng)用。本文以實(shí)際應(yīng)用為背景,專注于目標(biāo)檢測與跟蹤方法的研究。針對在固定背景(攝像頭固定)下檢測出運(yùn)動的目標(biāo)并實(shí)時(shí)進(jìn)行標(biāo)記跟蹤。本文在對傳統(tǒng)的跟蹤算法深入研究的基礎(chǔ)上,發(fā)現(xiàn)傳統(tǒng)的方法存在著一些問題,不能滿足在不同環(huán)境下對于檢測跟蹤的速度和精度上的要求。尤其是在復(fù)雜環(huán)境和多目標(biāo)環(huán)境下,跟蹤效果較差。因此,本文做了以下的工作:(1)在運(yùn)動目標(biāo)的檢測部分,傳統(tǒng)的檢測算法主要有幀間差分法、背景減除法和光流法。在充分分析對比了各種算法的優(yōu)缺點(diǎn)后,發(fā)現(xiàn)傳統(tǒng)的檢測算法在不同的環(huán)境下都不能實(shí)現(xiàn)很好的檢測效果。因此,本文在基于傳統(tǒng)的運(yùn)動目標(biāo)檢測算法的基礎(chǔ)上,實(shí)現(xiàn)了將背景減除法和背景建模相結(jié)合的目標(biāo)檢測算法進(jìn)行運(yùn)動目標(biāo)的檢測。在不同的環(huán)境下,實(shí)現(xiàn)了對運(yùn)動目標(biāo)比較理想的檢測效果。(2)在運(yùn)動目標(biāo)的跟蹤部分,如今使用最多的目標(biāo)跟蹤算法有MeanShift算法、粒子濾波算法以及Kalman算法。由于背景干擾、混亂、遮擋以及目標(biāo)快速移動,傳統(tǒng)的跟蹤算法存在著跟蹤漂移現(xiàn)象。因此,本文在深入研究粒子濾波跟蹤算法的基礎(chǔ)上,通過加入目標(biāo)的空間位置分布信息,進(jìn)行了改進(jìn),提出了基于空間位置--顏色直方圖的粒子濾波跟蹤算法。最后,通過大量的實(shí)驗(yàn)對比分析了不同算法的處理效果,驗(yàn)證了改進(jìn)后算法的有效性和魯棒性。對于現(xiàn)代智能化視頻監(jiān)控系統(tǒng)的發(fā)展及應(yīng)用有著重要的意義和實(shí)用價(jià)值。
[Abstract]:Video moving target detection and tracking is a hot research field of computer vision and pattern recognition in the intelligent video surveillance system, artificial intelligence, has been widely used in visual navigation and so on. Based on the practical application and research focus on the methods of detecting and tracking targets. Aiming at the fixed background (camera) detection the moving target and the real-time marking tracking. Based on the traditional tracking algorithm on the basis of the in-depth study found that the traditional method has some problems, can not meet the environment in different speed and accuracy for the detection of the tracking requirements. Especially in the complex environment and multi target environment, tracking effect is poor therefore, this paper has done the following work: (1) in the detection of the moving target, the traditional detection algorithms are mainly inter frame difference method, background subtraction and optical flow method. In the full analysis Comparing the advantages and disadvantages of various algorithms, found that the traditional detection algorithms in different environments can achieve good detection effect. Therefore, based on the traditional moving object detection algorithms on the detection method of background subtraction and background modeling combined target detection algorithm for moving target in a different environment, to achieve a better detection effect of the moving target. (2) in the moving target tracking part, now most of the target tracking algorithm with MeanShift algorithm, particle filter algorithm and Kalman algorithm. The background interference, chaos, fast moving target tracking and occlusion, the traditional algorithms exist the tracking drift phenomenon. Therefore, based on the in-depth study of the particle filter tracking algorithm, the spatial distribution of information to the target, has been improved, based on space Particle filter tracking algorithm -- color histogram. Finally, through experiments and analysis the effect of different algorithms, which verifies the effectiveness and robustness of the improved algorithm. The development and application of modern intelligent video monitoring system has important significance and practical value.

【學(xué)位授予單位】：沈陽航空航天大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前8條

1 李sチ，

本文編號：1357458