公告：凡是在
2025 年 4 月 15 日前註冊使用 Earth Engine 的非商業專案，都必須
驗證非商業用途資格，才能繼續存取 Earth Engine。
ee.Algorithms.TemporalSegmentation.LandTrendr
透過集合功能整理內容
你可以依據偏好儲存及分類內容。
根據 Landsat 偵測干擾和復原趨勢：擷取一段時間內的光譜變化軌跡，將時間序列圖像劃分為多個片段。系統會使用每張圖片的第一個波段尋找中斷點，並使用這些中斷點對所有後續波段執行擬合。系統會以 4 列的 2D 矩陣傳回中斷點，欄數則與圖片數量相同。前兩列是原始的 X 和 Y 值。第三列包含與預估區隔相符的 Y 值，第四列則包含 1 (如果對應點用做區隔頂點) 或 0 (如果不是)。輸出內容會以資料列的形式，顯示所有額外加裝的錶帶。中斷點擬合會假設值增加代表干擾，值減少代表復原。
請參閱：Kennedy, R.E.、Yang, Z. 和 Cohen, W.B.，2010 年。使用 Landsat 每年時間序列資料偵測森林干擾和復原趨勢：1. LandTrendr - 時間區隔演算法。Remote Sensing of Environment, 114(12), pp.2897-2910.
| 用量 | 傳回 |
|---|
ee.Algorithms.TemporalSegmentation.LandTrendr(timeSeries, maxSegments, spikeThreshold, vertexCountOvershoot, preventOneYearRecovery, recoveryThreshold, pvalThreshold, bestModelProportion, minObservationsNeeded) | 圖片 |
| 引數 | 類型 | 詳細資料 |
|---|
timeSeries | ImageCollection | 要從中擷取中斷點的年度時間序列。第一個頻帶用於尋找中斷點，後續所有頻帶都會使用這些中斷點進行調整。 |
maxSegments | 整數 | 要套用至時間序列的區隔數量上限。 |
spikeThreshold | 浮點值，預設值為 0.9 | 抑制尖峰的閾值 (1.0 表示不抑制)。 |
vertexCountOvershoot | 整數，預設值為 3 | 初始模型可能會超出 maxSegments + 1 個頂點，超出量為這個值。稍後，系統會將其修剪至 maxSegments + 1。 |
preventOneYearRecovery | 布林值，預設值為 false | 防止出現代表一年復原期的區隔。 |
recoveryThreshold | 浮點值，預設值為 0.25 | 如果區隔的復原率高於 1/recoveryThreshold (以年為單位)，則不允許該區隔。 |
pvalThreshold | 浮點值，預設值為 0.1 | 如果已調整模型的 p 值超過這個門檻，系統就會捨棄目前的模型，並使用 Levenberg-Marquardt 最佳化工具調整另一個模型。 |
bestModelProportion | 浮點值，預設值為 0.75 | 如果頂點較多的模型 p 值不超過最佳模型 p 值的 (2 - bestModelProportion) 倍，系統就會選擇這類模型。 |
minObservationsNeeded | 整數，預設值為 6 | 執行輸出擬合所需的最低觀察次數。 |
除非另有註明，否則本頁面中的內容是採用創用 CC 姓名標示 4.0 授權，程式碼範例則為阿帕契 2.0 授權。詳情請參閱《Google Developers 網站政策》。Java 是 Oracle 和/或其關聯企業的註冊商標。
上次更新時間：2025-07-26 (世界標準時間)。
[null,null,["上次更新時間：2025-07-26 (世界標準時間)。"],[[["\u003cp\u003eLandTrendr is a temporal segmentation algorithm designed to detect trends in disturbance and recovery within yearly Landsat time-series data.\u003c/p\u003e\n"],["\u003cp\u003eIt identifies breakpoints in spectral trajectories, using the first band of the image collection for initial detection and then fitting the breakpoints to all other bands.\u003c/p\u003e\n"],["\u003cp\u003eThese breakpoints, representing changes in land cover, are fitted to a model assuming increasing values indicate disturbance and decreasing values signify recovery.\u003c/p\u003e\n"],["\u003cp\u003eThe algorithm offers parameters for controlling spike dampening, segment recovery rates, model selection, and minimum data requirements to fine-tune the analysis.\u003c/p\u003e\n"],["\u003cp\u003eThe output is an image containing the original and fitted values, segment vertices, and optionally fitted values for additional bands.\u003c/p\u003e\n"]]],["The LandTrendr algorithm segments a time-series of images, using the first band to find breakpoints that identify spectral changes. These breakpoints are then applied to fit all other bands. Breakpoints are returned as a matrix, with the original data, fitted values, and segment vertex indicators. Increasing values suggest disturbance, while decreasing values suggest recovery. Users define parameters like maximum segments, spike dampening, and recovery rates to guide the fitting process. The algorithm outputs an image containing the results.\n"],null,["# ee.Algorithms.TemporalSegmentation.LandTrendr\n\nLandsat-based detection of Trends in Disturbance and Recovery: temporally segments a time-series of images by extracting the spectral trajectories of change over time. The first band of each image is used to find breakpoints, and those breakpoints are used to perform fitting on all subsequent bands. The breakpoints are returned as a 2-D matrix of 4 rows and as many columns as images. The first two rows are the original X and Y values. The third row contains the Y values fitted to the estimated segments, and the 4th row contains a 1 if the corresponding point was used as a segment vertex or 0 if not. Any additional fitted bands are appended as rows in the output. Breakpoint fitting assumes that increasing values represent disturbance and decreasing values represent recovery.\n\n\u003cbr /\u003e\n\nSee: Kennedy, R.E., Yang, Z. and Cohen, W.B., 2010. Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr - Temporal segmentation algorithms. Remote Sensing of Environment, 114(12), pp.2897-2910.\n\n| Usage | Returns |\n|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------|\n| `ee.Algorithms.TemporalSegmentation.LandTrendr(timeSeries, maxSegments, `*spikeThreshold* `, `*vertexCountOvershoot* `, `*preventOneYearRecovery* `, `*recoveryThreshold* `, `*pvalThreshold* `, `*bestModelProportion* `, `*minObservationsNeeded*`)` | Image |\n\n| Argument | Type | Details |\n|--------------------------|-------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------|\n| `timeSeries` | ImageCollection | Yearly time-series from which to extract breakpoints. The first band is usedto find breakpoints, and all subsequent bands are fitted using those breakpoints. |\n| `maxSegments` | Integer | Maximum number of segments to be fitted on the time series. |\n| `spikeThreshold` | Float, default: 0.9 | Threshold for dampening the spikes (1.0 means no dampening). |\n| `vertexCountOvershoot` | Integer, default: 3 | The initial model can overshoot the maxSegments + 1 vertices by this amount. Later, it will be pruned down to maxSegments + 1. |\n| `preventOneYearRecovery` | Boolean, default: false | Prevent segments that represent one year recoveries. |\n| `recoveryThreshold` | Float, default: 0.25 | If a segment has a recovery rate faster than 1/recoveryThreshold (in years), then the segment is disallowed. |\n| `pvalThreshold` | Float, default: 0.1 | If the p-value of the fitted model exceeds this threshold, then the current model is discarded and another one is fitted using the Levenberg-Marquardt optimizer. |\n| `bestModelProportion` | Float, default: 0.75 | Allows models with more vertices to be chosen if their p-value is no more than (2 - bestModelProportion) times the p-value of the best model. |\n| `minObservationsNeeded` | Integer, default: 6 | Min observations needed to perform output fitting. |"]]
