Algorism Research of Moving Object Detection and DSP Implementation

摘要：由于實(shí)際場(chǎng)景的多樣性，目前常用的運(yùn)動(dòng)目標(biāo)檢測(cè)算法都還存在一定程度的缺陷，因此本文提出了一種將幀差法和背景減法相結(jié)合的方法，實(shí)現(xiàn)快速精確地檢測(cè)和提取運(yùn)動(dòng)目標(biāo)。實(shí)驗(yàn)結(jié)果表明，本方法是比較實(shí)用的，能較好滿足實(shí)時(shí)視頻監(jiān)控系統(tǒng)的要求。最后將程序移植到基于DSP的平臺(tái)上，進(jìn)行相應(yīng)的優(yōu)化后基本滿足了實(shí)時(shí)性的要求。
關(guān)鍵詞：目標(biāo)檢測(cè)；幀差法；背景減法

Abstract: Because of the environment’s variety, the methods that have been used for moving object detection need to be improved. An algorithm based on two consecutive frames subtraction and background subtraction is presented and it can detect and extract object quickly and accurately. The results show that the proposed method is a practical one. It can meet the need of the real time video surveillance and monitoring system. The coding is transplanted in DSP, and the project is executed successfully on CCS simulator.
Keywords: Object detection; Frames subtraction; Background subtraction
引言
運(yùn)動(dòng)目標(biāo)的檢測(cè)在智能監(jiān)控等領(lǐng)域中得到了廣泛的應(yīng)用。運(yùn)動(dòng)目標(biāo)的檢測(cè)就是從視頻流中去除靜止背景提取出運(yùn)動(dòng)的目標(biāo)，運(yùn)動(dòng)目標(biāo)的有效分割對(duì)跟蹤等后期處理非常關(guān)鍵。
本文提出了將幀間差分和背景減法相結(jié)合的方法。首先選取一幀作為背景幀, 建立各像素點(diǎn)的高斯模型。再運(yùn)用幀間差分法對(duì)相鄰兩幀圖像進(jìn)行差分處理, 區(qū)分出背景點(diǎn)和變化的區(qū)域。然后將變化區(qū)域與背景幀的對(duì)應(yīng)區(qū)域進(jìn)行模型擬合區(qū)分出顯露區(qū)和運(yùn)動(dòng)物體。
1 運(yùn)動(dòng)目標(biāo)檢測(cè)算法總體流程
采用幀間差分與背景減法相結(jié)合的算法進(jìn)行運(yùn)動(dòng)目標(biāo)檢測(cè)，包括運(yùn)動(dòng)目標(biāo)的檢測(cè)和將檢測(cè)到的運(yùn)動(dòng)目標(biāo)從背景中分割出來(lái)兩部分，其系統(tǒng)框架流程圖如圖1所示。