本文選題：遙感圖像分割　切入點(diǎn)：模糊C均值　出處：《中國(guó)圖象圖形學(xué)報(bào)》2017年07期

【摘要】：目的傳統(tǒng)FCM算法及其改進(jìn)算法均只采用隸屬度作為分割判據(jù)實(shí)現(xiàn)圖像分割。然而,在分割過(guò)程中聚類(lèi)中心易受到同質(zhì)區(qū)域內(nèi)幾何噪聲的影響,導(dǎo)致此類(lèi)算法難以有效分割具有幾何噪聲的圖像。為了解決這一類(lèi)問(wèn)題,提出一種利用包含度和隸屬度的遙感影像模糊分割算法。方法該算法假設(shè)同一聚類(lèi)對(duì)每個(gè)像素都有不同程度的包含度,將包含度作為一種新測(cè)度來(lái)描述聚類(lèi)與像素間關(guān)系,并將包含度納入目標(biāo)函數(shù)中。該算法通過(guò)迭代最小化目標(biāo)函數(shù)來(lái)得到最優(yōu)的隸屬度和包含度,然后,通過(guò)反模糊化隸屬度和包含度之積實(shí)現(xiàn)帶有幾何噪聲的遙感圖像的分割。結(jié)果采用本文算法分別對(duì)模擬圖像,真實(shí)遙感影像進(jìn)行分割實(shí)驗(yàn),并與FCM算法和FLICM算法進(jìn)行對(duì)比,定性結(jié)果表明,對(duì)含有幾何噪聲的區(qū)域,提出算法的用戶精度和產(chǎn)品精度均高于FCM算法和FLICM算法,且總精度和Kappa值也高于對(duì)比算法。實(shí)驗(yàn)結(jié)果表明,本文算法能夠抵抗幾何噪聲對(duì)圖像分割的影響,且分割精度遠(yuǎn)遠(yuǎn)高于其他兩種算法的分割精度。結(jié)論提出算法通過(guò)考慮聚類(lèi)對(duì)像素的包含性,能夠有效抵抗幾何噪聲對(duì)圖像分割的影響,使得算法具有較高的抗幾何噪聲能力,進(jìn)而提高該算法對(duì)含有幾何噪聲圖像的分割精度。提出算法適用于包含幾何噪聲的高分辨率遙感圖像,具有很好的抗幾何噪聲性。
[Abstract]:Objective the traditional FCM algorithm and its improved algorithm only use membership degree as the segmentation criterion to realize image segmentation. However, the clustering center is easily affected by the geometric noise in the homogeneous region during the segmentation process. In order to solve this kind of problem, it is difficult to segment images with geometric noise effectively. A fuzzy segmentation algorithm for remote sensing image using inclusion degree and membership degree is proposed. The algorithm assumes that the same clustering has different degrees of inclusion for each pixel, and uses inclusion degree as a new measure to describe the relationship between clustering and pixels. And the inclusion degree is incorporated into the objective function. The optimal membership degree and inclusion degree are obtained by iterative minimization of the objective function, and then, The segmentation of remote sensing image with geometric noise is realized by using the product of defuzzification membership degree and inclusion degree. Results the proposed algorithm is used to segment simulated image and real remote sensing image separately, and compared with FCM algorithm and FLICM algorithm. The qualitative results show that the user accuracy and product precision of the proposed algorithm are higher than those of FCM and FLICM algorithms, and the total accuracy and Kappa value of the proposed algorithm are also higher than those of the contrast algorithm for the regions with geometric noise. The proposed algorithm can resist the influence of geometric noise on image segmentation, and the segmentation accuracy is much higher than that of the other two algorithms. It can effectively resist the influence of geometric noise on image segmentation, so that the algorithm has a higher ability to resist geometric noise. The proposed algorithm is suitable for high resolution remote sensing images with geometric noise and has good geometric noise resistance.
【作者單位】： 遼寧工程技術(shù)大學(xué)測(cè)繪與地理科學(xué)學(xué)院;
【基金】：國(guó)家自然科學(xué)基金項(xiàng)目(41301479,41271435) 遼寧省自然科學(xué)基金項(xiàng)目(2015020090)~~
【分類(lèi)號(hào)】：TP751
，

本文編號(hào)：1697525