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。
縮減 ImageCollection
透過集合功能整理內容
你可以依據偏好儲存及分類內容。
如要在 ImageCollection
中合成圖片,請使用 imageCollection.reduce()
。這會將集合中的所有圖片合成單一圖片,用來代表圖片的極小值、極大值、平均值或標準差。(如要進一步瞭解 Reducer,請參閱「Reducer 一節」)。例如,如要從集合建立中位值圖片:
程式碼編輯器 (JavaScript)
// Load a Landsat 8 collection for a single path-row.
var collection = ee.ImageCollection('LANDSAT/LC08/C02/T1_TOA')
.filter(ee.Filter.eq('WRS_PATH', 44))
.filter(ee.Filter.eq('WRS_ROW', 34))
.filterDate('2014-01-01', '2015-01-01');
// Compute a median image and display.
var median = collection.median();
Map.setCenter(-122.3578, 37.7726, 12);
Map.addLayer(median, {bands: ['B4', 'B3', 'B2'], max: 0.3}, 'Median');
Python 設定
請參閱「
Python 環境」頁面,瞭解 Python API 和如何使用 geemap
進行互動式開發。
import ee
import geemap.core as geemap
Colab (Python)
# Load a Landsat 8 collection for a single path-row.
collection = (
ee.ImageCollection('LANDSAT/LC08/C02/T1_TOA')
.filter(ee.Filter.eq('WRS_PATH', 44))
.filter(ee.Filter.eq('WRS_ROW', 34))
.filterDate('2014-01-01', '2015-01-01')
)
# Compute a median image and display.
median = collection.median()
m = geemap.Map()
m.set_center(-122.3578, 37.7726, 12)
m.add_layer(median, {'bands': ['B4', 'B3', 'B2'], 'max': 0.3}, 'Median')
m
在輸出圖像的每個位置和每個頻帶中,像素值都是輸入圖像 (集合中的圖像) 中所有未遮罩像素的中位數。在前述範例中,median()
是下列呼叫的便利方法:
程式碼編輯器 (JavaScript)
// Reduce the collection with a median reducer.
var median = collection.reduce(ee.Reducer.median());
// Display the median image.
Map.addLayer(median,
{bands: ['B4_median', 'B3_median', 'B2_median'], max: 0.3},
'Also median');
Python 設定
請參閱「
Python 環境」頁面,瞭解 Python API 和如何使用 geemap
進行互動式開發。
import ee
import geemap.core as geemap
Colab (Python)
# Reduce the collection with a median reducer.
median = collection.reduce(ee.Reducer.median())
# Display the median image.
m.add_layer(
median,
{'bands': ['B4_median', 'B3_median', 'B2_median'], 'max': 0.3},
'Also median',
)
m
請注意,由於使用 reduce()
而非便利方法,因此頻道名稱有所不同。具體來說,縮減器的名稱已附加至頻道名稱。
您也可以使用 reduce()
進行更複雜的縮減作業。舉例來說,如要計算集合中的長期線性趨勢,請使用其中一個線性迴歸縮減器。下列程式碼會計算 MODIS 增強植被指數 (EVI) 的線性趨勢:
程式碼編輯器 (JavaScript)
// This function adds a band representing the image timestamp.
var addTime = function(image) {
return image.addBands(image.metadata('system:time_start')
// Convert milliseconds from epoch to years to aid in
// interpretation of the following trend calculation.
.divide(1000 * 60 * 60 * 24 * 365));
};
// Load a MODIS collection, filter to several years of 16 day mosaics,
// and map the time band function over it.
var collection = ee.ImageCollection('MODIS/006/MYD13A1')
.filterDate('2004-01-01', '2010-10-31')
.map(addTime);
// Select the bands to model with the independent variable first.
var trend = collection.select(['system:time_start', 'EVI'])
// Compute the linear trend over time.
.reduce(ee.Reducer.linearFit());
// Display the trend with increasing slopes in green, decreasing in red.
Map.setCenter(-96.943, 39.436, 5);
Map.addLayer(
trend,
{min: 0, max: [-100, 100, 10000], bands: ['scale', 'scale', 'offset']},
'EVI trend');
Python 設定
請參閱「
Python 環境」頁面,瞭解 Python API 和如何使用 geemap
進行互動式開發。
import ee
import geemap.core as geemap
Colab (Python)
# This function adds a band representing the image timestamp.
def add_time(image):
return image.addBands(
image.metadata('system:time_start')
# Convert milliseconds from epoch to years to aid in
# interpretation of the following trend calculation.
.divide(1000 * 60 * 60 * 24 * 365)
)
# Load a MODIS collection, filter to several years of 16 day mosaics,
# and map the time band function over it.
collection = (
ee.ImageCollection('MODIS/006/MYD13A1')
.filterDate('2004-01-01', '2010-10-31')
.map(add_time)
)
# Select the bands to model with the independent variable first.
trend = collection.select(['system:time_start', 'EVI']).reduce(
# Compute the linear trend over time.
ee.Reducer.linearFit()
)
# Display the trend with increasing slopes in green, decreasing in red.
m.set_center(-96.943, 39.436, 5)
m = geemap.Map()
m.add_layer(
trend,
{
'min': 0,
'max': [-100, 100, 10000],
'bands': ['scale', 'scale', 'offset'],
},
'EVI trend',
)
m
請注意,本例中縮減作業的輸出內容是兩個帶狀圖像,其中一個帶狀圖像代表線性迴歸的斜率 (scale
),另一個帶狀圖像代表截距 (offset
)。請參閱 API 說明文件,查看可用縮減器的清單,以便將 ImageCollection
縮減為單一 Image
。
複合物沒有投影
透過縮減圖片集合所建立的複合圖片,可產生任何要求的專案中的像素,因此沒有固定的輸出投影方式。相反地,合成圖層具有 1 度解析度像素的 預設投影方式 WGS-84。使用預設投影的組合會在要求的任何輸出投影中計算。您可以在程式碼編輯器中顯示複合圖 (瞭解程式碼編輯器如何設定縮放和投影),或是在匯總 (例如 ReduceRegion
或 Export
) 中明確指定投影/縮放。
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:2025-07-25 (世界標準時間)。
[null,null,["上次更新時間:2025-07-25 (世界標準時間)。"],[[["\u003cp\u003eUse \u003ccode\u003eimageCollection.reduce()\u003c/code\u003e to composite images in an \u003ccode\u003eImageCollection\u003c/code\u003e into a single image representing a statistical summary (e.g., median, mean) of the collection.\u003c/p\u003e\n"],["\u003cp\u003eThe \u003ccode\u003ereduce()\u003c/code\u003e function utilizes reducers like \u003ccode\u003eee.Reducer.median()\u003c/code\u003e to calculate the desired composite, with band names reflecting the reducer used.\u003c/p\u003e\n"],["\u003cp\u003eMore complex reductions, such as calculating linear trends, are possible using specific reducers like \u003ccode\u003eee.Reducer.linearFit()\u003c/code\u003e.\u003c/p\u003e\n"],["\u003cp\u003eComposite images generated from reducing an image collection do not have a fixed projection and will be computed based on the requested output projection.\u003c/p\u003e\n"]]],[],null,["# Reducing an ImageCollection\n\nTo composite images in an `ImageCollection`, use\n`imageCollection.reduce()`. This will composite all the images in the\ncollection to a single image representing, for example, the min, max, mean or standard\ndeviation of the images.\n(See the [Reducers section](/earth-engine/guides/reducers_image_collection)\nfor more information about reducers). For example, to create a median value image from a\ncollection:\n\n### Code Editor (JavaScript)\n\n```javascript\n// Load a Landsat 8 collection for a single path-row.\nvar collection = ee.ImageCollection('LANDSAT/LC08/C02/T1_TOA')\n .filter(ee.Filter.eq('WRS_PATH', 44))\n .filter(ee.Filter.eq('WRS_ROW', 34))\n .filterDate('2014-01-01', '2015-01-01');\n\n// Compute a median image and display.\nvar median = collection.median();\nMap.setCenter(-122.3578, 37.7726, 12);\nMap.addLayer(median, {bands: ['B4', 'B3', 'B2'], max: 0.3}, 'Median');\n```\nPython setup\n\nSee the [Python Environment](/earth-engine/guides/python_install) page for information on the Python API and using\n`geemap` for interactive development. \n\n```python\nimport ee\nimport geemap.core as geemap\n```\n\n### Colab (Python)\n\n```python\n# Load a Landsat 8 collection for a single path-row.\ncollection = (\n ee.ImageCollection('LANDSAT/LC08/C02/T1_TOA')\n .filter(ee.Filter.eq('WRS_PATH', 44))\n .filter(ee.Filter.eq('WRS_ROW', 34))\n .filterDate('2014-01-01', '2015-01-01')\n)\n\n# Compute a median image and display.\nmedian = collection.median()\nm = geemap.Map()\nm.set_center(-122.3578, 37.7726, 12)\nm.add_layer(median, {'bands': ['B4', 'B3', 'B2'], 'max': 0.3}, 'Median')\nm\n```\n\nAt each location in the output image, in each band, the pixel value is the median of all\nunmasked pixels in the input imagery (the images in the collection). In the previous\nexample, `median()` is a convenience method for the following call:\n\n### Code Editor (JavaScript)\n\n```javascript\n// Reduce the collection with a median reducer.\nvar median = collection.reduce(ee.Reducer.median());\n\n// Display the median image.\nMap.addLayer(median,\n {bands: ['B4_median', 'B3_median', 'B2_median'], max: 0.3},\n 'Also median');\n```\nPython setup\n\nSee the [Python Environment](/earth-engine/guides/python_install) page for information on the Python API and using\n`geemap` for interactive development. \n\n```python\nimport ee\nimport geemap.core as geemap\n```\n\n### Colab (Python)\n\n```python\n# Reduce the collection with a median reducer.\nmedian = collection.reduce(ee.Reducer.median())\n\n# Display the median image.\nm.add_layer(\n median,\n {'bands': ['B4_median', 'B3_median', 'B2_median'], 'max': 0.3},\n 'Also median',\n)\nm\n```\n\nNote that the band names differ as a result of using `reduce()` instead of the\nconvenience method. Specifically, the names of the reducer have been appended to the\nband names.\n\nMore complex reductions are also possible using `reduce()`. For\nexample, to compute the long term linear trend over a collection, use one of the linear\nregression reducers. The following code computes the linear trend of MODIS Enhanced\nVegetation Index (EVI):\n\n### Code Editor (JavaScript)\n\n```javascript\n// This function adds a band representing the image timestamp.\nvar addTime = function(image) {\n return image.addBands(image.metadata('system:time_start')\n // Convert milliseconds from epoch to years to aid in\n // interpretation of the following trend calculation.\n .divide(1000 * 60 * 60 * 24 * 365));\n};\n\n// Load a MODIS collection, filter to several years of 16 day mosaics,\n// and map the time band function over it.\nvar collection = ee.ImageCollection('MODIS/006/MYD13A1')\n .filterDate('2004-01-01', '2010-10-31')\n .map(addTime);\n\n// Select the bands to model with the independent variable first.\nvar trend = collection.select(['system:time_start', 'EVI'])\n // Compute the linear trend over time.\n .reduce(ee.Reducer.linearFit());\n\n// Display the trend with increasing slopes in green, decreasing in red.\nMap.setCenter(-96.943, 39.436, 5);\nMap.addLayer(\n trend,\n {min: 0, max: [-100, 100, 10000], bands: ['scale', 'scale', 'offset']},\n 'EVI trend');\n```\nPython setup\n\nSee the [Python Environment](/earth-engine/guides/python_install) page for information on the Python API and using\n`geemap` for interactive development. \n\n```python\nimport ee\nimport geemap.core as geemap\n```\n\n### Colab (Python)\n\n```python\n# This function adds a band representing the image timestamp.\ndef add_time(image):\n return image.addBands(\n image.metadata('system:time_start')\n # Convert milliseconds from epoch to years to aid in\n # interpretation of the following trend calculation.\n .divide(1000 * 60 * 60 * 24 * 365)\n )\n\n\n# Load a MODIS collection, filter to several years of 16 day mosaics,\n# and map the time band function over it.\ncollection = (\n ee.ImageCollection('MODIS/006/MYD13A1')\n .filterDate('2004-01-01', '2010-10-31')\n .map(add_time)\n)\n\n# Select the bands to model with the independent variable first.\ntrend = collection.select(['system:time_start', 'EVI']).reduce(\n # Compute the linear trend over time.\n ee.Reducer.linearFit()\n)\n\n# Display the trend with increasing slopes in green, decreasing in red.\nm.set_center(-96.943, 39.436, 5)\nm = geemap.Map()\nm.add_layer(\n trend,\n {\n 'min': 0,\n 'max': [-100, 100, 10000],\n 'bands': ['scale', 'scale', 'offset'],\n },\n 'EVI trend',\n)\nm\n```\n\nNote that the output of the reduction in this example is a two banded image\nwith one band for the slope of a linear regression (`scale`) and one band\nfor the intercept (`offset`). Explore the API documentation to see a list of\nthe reducers that are available to reduce an `ImageCollection` to a single\n`Image`.\n\nComposites have no projection\n-----------------------------\n\nComposite images created by reducing an image collection are able to produce pixels\nin any requested projection and therefore *have no fixed output projection* .\nInstead, composites have\n[the default\nprojection](/earth-engine/guides/projections#the-default-projection) of WGS-84 with 1-degree resolution pixels. Composites with the default\nprojection will be computed in whatever output projection is requested. A request\noccurs by displaying the composite in the Code Editor (learn about how the Code editor\nsets [scale](/earth-engine/guides/scale#scale-of-analysis) and\n[projection](/earth-engine/guides/projections)), or by explicitly specifying a\nprojection/scale as in an aggregation such as\n`ReduceRegion` or `Export`."]]