【摘要】：制種玉米田在高空間分辨率遙感影像上呈現(xiàn)的明顯條帶狀紋理,是有效區(qū)分光譜值相近的大田玉米和制種玉米的重要信息。該文在新疆維吾爾自治區(qū)奇臺縣玉米種植區(qū)以高空間分辨率的無人機遙感影像為數(shù)據(jù)源,針對制種玉米識別的紋理特征計算尺度問題,首先采用最近鄰內(nèi)插法對制種玉米和大田玉米樣本田塊的無人機影像進行重采樣,得到不同分辨率的樣本;然后用融合Uniform-LBP(local binary pattern)和GLCM(gray level co-occurrence matrix)方法得到提取玉米田塊紋理特征合理GLCM參數(shù),其中方向參數(shù)為0°、45°、90°和135°這4個方向上的紋理特征值的平均值、距離為5~7像元、灰度級為8;通過多尺度對比分析,得到最適宜區(qū)分制種玉米與大田玉米的紋理辨率為0.6~0.9 m。最后采用奇臺縣的0.7 m分辨率的Kompsat-3遙感影像進行驗證,在多時相EVI(enhanced vegetation index)光譜信息識別玉米的基礎(chǔ)上,利用本文確定的紋理分析方法,通過決策樹建立規(guī)則識別制種玉米,識別精度達90.9%。通過該文的研究,可為高空間分辨率遙感制種玉米田監(jiān)管提供支撐。
[Abstract]:The obvious banded texture of seed production corn field on the remote sensing image of high altitude resolution is an important information for effectively distinguishing field corn with similar spectral value from seed production corn. In this paper, the UAV remote sensing image with high spatial resolution is used as the data source in the corn growing area of Qitai County, Xinjiang Uygur Autonomous region. Aiming at the texture feature calculation scale of seed corn recognition, the UAV image of seed production corn and field corn sample field is resampled by nearest neighbor interpolation method, and the samples with different resolutions are obtained. Then the reasonable GLCM parameters of extracting texture features of corn field were obtained by fusion Uniform-LBP (local binary pattern) and GLCM (gray level co-occurrence matrix), in which the average value of texture eigenvalues in four directions was 0 擄, 45 擄, 90 擄and 135 擄, the distance was 5 鈮，

本文編號：2515131