本文選題：視覺注意　切入點(diǎn)：水平集方法　出處：《紅外》2016年11期

【摘要】：針對(duì)傳統(tǒng)紅外目標(biāo)檢測(cè)與識(shí)別方法所存在的問題,即其處理過程總是盲目地對(duì)全圖進(jìn)行耗時(shí)搜索,提出了一種基于視覺注意機(jī)制和水平集方法的紅外海面目標(biāo)檢測(cè)與識(shí)別方法。首先,搜索原始圖像中的顯著性區(qū)域,并以獲勝點(diǎn)的形式表示它們。接著,基于所得到的顯著性區(qū)域,自動(dòng)初始化水平集函數(shù),并使演化過程朝著期望的目標(biāo)輪廓方向挺進(jìn),直至演化過程到達(dá)最終的平衡狀態(tài)。最后,針對(duì)遠(yuǎn)距離(近距離)成像時(shí)的輸入數(shù)據(jù),給出檢測(cè)結(jié)果(基于不變矩和神經(jīng)網(wǎng)絡(luò)框架的識(shí)別結(jié)果)。對(duì)真實(shí)紅外海面目標(biāo)進(jìn)行的實(shí)驗(yàn)證實(shí)了本文方法的有效性。
[Abstract]:Aiming at the problems of traditional infrared target detection and recognition methods, that is, the processing process always blindly carries out the time-consuming search of the whole image. In this paper, a method of infrared sea surface target detection and recognition based on visual attention mechanism and level set method is proposed. Firstly, the salient regions in the original image are searched, and they are represented in the form of winning points. Based on the obtained salience region, the level set function is automatically initialized, and the evolution process moves towards the desired target profile until the evolution process reaches the final equilibrium state. For the input data of long range (short range) imaging, the detection results (based on the invariant moment and neural network framework) are given. Experiments on the real infrared sea surface target show the effectiveness of the proposed method.
【作者單位】： 93501部隊(duì)17分隊(duì);電子科技大學(xué)航空航天學(xué)院;空軍第一航空學(xué)院航空彈藥教研室;
【基金】：中央高校基本科研業(yè)務(wù)費(fèi)(ZYGX2015KYQD032)
【分類號(hào)】：TP391.41
，

本文編號(hào)：1660581