From April-May 2020, Sentinel-Hub had organized the third edition of their custom script competition. The competition was organized in collaboration with the Copernicus EU Earth Observation programme, the European Space Agency and AI4EO consortium.

從2020年4月至5月，Sentinel-Hub組織了第三次自定義腳本競(jìng)賽 。 該競(jìng)賽是與哥白尼歐盟地球觀測(cè)計(jì)劃 ， 歐洲航天局和AI4EO財(cái)團(tuán)合作組織的 。

Motive of the competition:

比賽動(dòng)機(jī)：

To look for new, innovative scripts, which enable users to make sense of Earth Observation data. It was designed to find solutions to the huge challenges in the new ‘normal’

尋找新的，創(chuàng)新的腳本，使用戶(hù)能夠理解地球觀測(cè)數(shù)據(jù)。 它旨在為新的“ 正常 ”情況下的巨大挑戰(zhàn)找到解決方案

What was the objective of our submission:

我們提交的目標(biāo)是什么 ：

Covid-19 has led to many world governments enforcing emergency quarantine measures. What is the impact of such policies on the environment? Can we measure the change in pollution levels? Is there a co-relation to economic activity? And can we build leading indicators which enable us to measure economic activity at a global scale. Through our project, we tried to analyze satellite imagery to identify the answers.

Covid-19導(dǎo)致許多世界政府強(qiáng)制執(zhí)行緊急隔離措施。 此類(lèi)政策對(duì)環(huán)境有何影響？ 我們可以衡量污染水平的變化嗎？ 與經(jīng)濟(jì)活動(dòng)是否有關(guān)聯(lián)？ 我們是否可以建立領(lǐng)先的指標(biāo)，使我們能夠在全球范圍內(nèi)衡量經(jīng)濟(jì)活動(dòng)。 通過(guò)我們的項(xiàng)目，我們嘗試分析衛(wèi)星圖像以找出答案。

Tools used:

使用的工具：

The tools make the man, or so they say!

這些工具造就了男人，或者他們說(shuō)！

I always tell my team to first start, by researching the available tools. Because the right tools can make or break a project!

我總是告訴我的團(tuán)隊(duì)首先研究可用的工具。 因?yàn)檎_的工具可以建立或破壞項(xiàng)目！

The tools that were available for this project were:

該項(xiàng)目可用的工具有：

Sentinel-Hub EO Browser

Sentinel-Hub EO瀏覽器

Sentinel-Hub API access

Sentinel-Hub API訪問(wèn)

Picterra’s AI platform

Picterra的AI平臺(tái)

euroData Cube

euroData多維數(shù)據(jù)集

We first used Sentinel-5P data to plot NO2 levels. This was done to identify any anomalies (sudden changes) in the data. This was then collaborated with VHR imagery (Very high resolution imagery)

我們首先使用Sentinel-5P數(shù)據(jù)繪制NO2水平。 這樣做是為了識(shí)別數(shù)據(jù)中的任何異常(突然變化)。 然后與VHR影像(非常高分辨率的影像)合作

Statistical analysis of this data was used to show a significant difference in countries like Germany (where the lockdown was strictly implemented) vs countries like Italy.

通過(guò)對(duì)該數(shù)據(jù)進(jìn)行統(tǒng)計(jì)分析，可以發(fā)現(xiàn)德國(guó)(嚴(yán)格執(zhí)行鎖定)等國(guó)家與意大利等國(guó)家之間存在顯著差異。

Plot of NO2 levels for Germany & Italy (note the presence of missing values) Source: Sentinel-5P processed data德國(guó)和意大利的NO2水平圖(注意存在遺漏值)來(lái)源：Sentinel-5P處理的數(shù)據(jù)

I have uploaded the notebook used for plotting the above plot on euroData Cube’s Contribution page. you can check it via the link given below:

我已經(jīng)將用于繪制上述圖表的筆記本上載到euroData Cube的“貢獻(xiàn)”頁(yè)面。 您可以通過(guò)以下鏈接進(jìn)行檢查：

https://eurodatacube.com/marketplace/notebooks/contributions/NO2_Analysis_Covid19_Lockdowns.ipynb

https://eurodatacube.com/marketplace/notebooks/contributions/NO2_Analysis_Covid19_Lockdowns.ipynb

NO2 is produced by vehicular traffic and as un-burnt residue from chemical processes. This can be collaborated by counting cars from high resolution satellite imagery.

機(jī)動(dòng)車(chē)交通產(chǎn)生的二氧化氮是化學(xué)過(guò)程中未燃燒的殘留物。 可以通過(guò)計(jì)算高分辨率衛(wèi)星圖像中的汽車(chē)來(lái)進(jìn)行協(xié)作。

We will now use Picterra’s machine learning platform to identify vehicles in VHR imagery.

現(xiàn)在，我們將使用Picterra的機(jī)器學(xué)習(xí)平臺(tái)來(lái)識(shí)別VHR圖像中的車(chē)輛。

We need to first login to Picterra’s geo-spatial imagery platform. They have a free trial version, which you can use (without providing any credit cards!)

我們需要先登錄Picterra的地理空間圖像平臺(tái)。 他們有一個(gè)免費(fèi)的試用版，您可以使用它( 無(wú)需提供任何信用卡！ )

Once on the home page, you can Create a new project

進(jìn)入主頁(yè)后，您可以創(chuàng)建一個(gè)新項(xiàng)目

Now you can upload images from your local drive (other options are Online map imagery OR the option to Buy satellite images). Once uploaded, you can then use it to train a detector.

現(xiàn)在，您可以從本地驅(qū)動(dòng)器上載圖像 (其他選項(xiàng)是“ 在線地圖圖像”或“ 購(gòu)買(mǎi)衛(wèi)星圖像 ”選項(xiàng))。 上傳后，您可以使用它來(lái)訓(xùn)練探測(cè)器。

Once the images are uploaded, we then go to the training section. Here we can use the platform to train a detector. However there are pre-built detectors which you can use. But for this demo, we will train a new one

圖片上傳后，我們將轉(zhuǎn)到培訓(xùn)部分。 在這里，我們可以使用平臺(tái)訓(xùn)練探測(cè)器。 但是，您可以使用預(yù)制的探測(cè)器。 但是對(duì)于這個(gè)演示，我們將訓(xùn)練一個(gè)新的

Training mode!訓(xùn)練模式！ Available pre-trained models可用的預(yù)訓(xùn)練模型

Click on the Train a new detector

單擊訓(xùn)練新的檢測(cè)器

There are some custom base models. We chose the Vehicles option

有一些自定義基本模型。 我們選擇了車(chē)輛選項(xiàng)

We want to detect cars

我們要檢測(cè)汽車(chē)

There are two available detection type: i) Count and ii) Segmentation

有兩種可用的檢測(cè)類(lèi)型：i)計(jì)數(shù)和ii)分段

We select the default Count option. and then press the Create button

我們選擇默認(rèn)的“ 計(jì)數(shù)”選項(xiàng)。 然后按創(chuàng)建按鈕

Then add the images that you had added to the project. and press the Start Training button

然后添加您已添加到項(xiàng)目中的圖像。 然后按開(kāi)始訓(xùn)練按鈕

When you do it for the first time, Picterra gives you a nice video tutorial that walks you through the training process.

第一次進(jìn)行此操作時(shí)，Picterra會(huì)為您提供一個(gè)不錯(cuò)的視頻教程，引導(dǎo)您完成培訓(xùn)過(guò)程。

The interface is very intuitive, and once you understand the controls, it becomes easy to navigate.

界面非常直觀，一旦您了解了控件，就可以輕松瀏覽。

First select the Training button to the left. After that you can select the area of interest, as marked by the yellow square. Clicking on the selected AOI, will zoom into the image

首先選擇左側(cè)的訓(xùn)練按鈕。 之后，您可以選擇感興趣的區(qū)域，以黃色正方形標(biāo)記。 單擊所選的AOI，將放大圖像

Now click on the Polygon button and select the Circle option

現(xiàn)在單擊“ 多邊形”按鈕，然后選擇“ 圓”選項(xiàng)

You can now click on each of the parked cars within the area of interest and draw a circle on top. This is now your annotated train dataset. You could have selected the polygon option and drawn a box across the car outline, but I am too lazy for that!

現(xiàn)在，您可以單擊感興趣區(qū)域內(nèi)的每個(gè)停放的汽車(chē)，并在頂部繪制一個(gè)圓圈。 現(xiàn)在這是您帶注釋的火車(chē)數(shù)據(jù)集。 您可以選擇“多邊形”選項(xiàng)，并在汽車(chē)輪廓上繪制一個(gè)框，但是我實(shí)在太懶了！

Now select the Testing button on the left menu and click and mark the testing area on the image. You can annotate the testing dataset. But here I will skip that step — Just to check what happens!

現(xiàn)在，選擇左側(cè)菜單上的“ 測(cè)試”按鈕，然后單擊并在圖像上標(biāo)記測(cè)試區(qū)域。 您可以注釋測(cè)試數(shù)據(jù)集。 但是在這里，我將跳過(guò)該步驟- 僅檢查發(fā)生了什么！

Now click on the Train Detector button at the top of the dashboard.

現(xiàn)在，單擊儀表板頂部的“ 火車(chē)檢測(cè)器”按鈕。

Now you can go and have a coffee ( I prefer my masala chai!)

現(xiàn)在您可以去喝咖啡了( 我更喜歡我的咖喱柴！ )

https://en.wikipedia.org/wiki/Masala_chai#/media/File:Masala_Chai.JPGhttps://en.wikipedia.org/wiki/Masala_chai#/media/File:Masala_Chai.JPG

OR/AND

或/與

can check the educational resources provided. Do check the Picterra University. It has some good resources on end-to-end machine learning for geo-spatial images.

可以查看提供的教育資源。 請(qǐng)檢查Picterra大學(xué)。 它在地理空間圖像的端到端機(jī)器學(xué)習(xí)方面有一些不錯(cuò)的資源。

Since the training is now finished, let's go and check the performance of the detector.

由于培訓(xùn)現(xiàn)已結(jié)束，因此我們開(kāi)始檢查檢測(cè)器的性能。

Testing output測(cè)試輸出

The performance is good…but not great! At the bottom right, it has mis-identified the roof of the warehouse as a car.

表現(xiàn)不錯(cuò)…但不是很好！ 在右下角，它錯(cuò)誤地將倉(cāng)庫(kù)的屋頂標(biāo)識(shí)為汽車(chē)。

For a detector its important to know what it should not consider as an object. In machine learning terms - false positives should be low.

對(duì)于檢測(cè)器來(lái)說(shuō)，重要的是要知道它不應(yīng)該考慮什么。 用機(jī)器學(xué)習(xí)的術(shù)語(yǔ)來(lái)說(shuō)，誤報(bào)率應(yīng)該很低。

So we will annotate some more vehicles for our training set. We select a larger area near our previous training set.

因此，我們將為訓(xùn)練集注釋更多的車(chē)輛。 我們?cè)谥暗挠?xùn)練集附近選擇了一個(gè)較大的區(qū)域。

Post training, I get a warning: it seems I have some overlapping annotations in the training set. which is true. This can cause mergers while running the inference jobs and give an incorrect count.

訓(xùn)練后，我得到一個(gè)警告：看來(lái)我在訓(xùn)練集中有一些重疊的注釋。 沒(méi)錯(cuò) 這可能會(huì)在運(yùn)行推理作業(yè)時(shí)導(dǎo)致合并，并且計(jì)數(shù)不正確。

So let’s correct that -

所以讓我們更正-

Use the Select button on the left to select any of the annotations & modify the marker (green circle).

使用左側(cè)的“選擇”按鈕選擇任何注釋并修改標(biāo)記( 綠色圓圈 )。

Modifying the training data to avoid merged annotations修改訓(xùn)練數(shù)據(jù)以避免合并注釋

Retraining the detector, we get the following output.

重新訓(xùn)練檢測(cè)器，我們得到以下輸出。

Testing area after re-training重新訓(xùn)練后的測(cè)試區(qū)域

Seems good. We have missed one large white vehicle at the left, but now the misclassification from the roof is missing. And it has got the car to the extreme left corner.

看起來(lái)不錯(cuò)。 我們錯(cuò)過(guò)了左側(cè)的一輛大型白色車(chē)輛，但是現(xiàn)在缺少了從車(chē)頂分類(lèi)錯(cuò)誤的信息。 它已將汽車(chē)推到了最左端。

The detector can be trained with more data. Let’s now run the detector on the entire image to see the evaluation performance. We will also evaluate it on a new image.

可以用更多數(shù)據(jù)訓(xùn)練檢測(cè)器。 現(xiàn)在讓我們?cè)谡麄€(gè)圖像上運(yùn)行檢測(cè)器以查看評(píng)估性能。 我們還將在新圖像上對(duì)其進(jìn)行評(píng)估。

Added new un-seen images & the trained detector添加了新的看不見(jiàn)的圖像和訓(xùn)練有素的探測(cè)器

Press the Run Detector button in front of the image. A prompt informs you about the number of processing quota required for running the data. Since we are running on a free-quota, we press Start Detection.

按下圖像前面的“運(yùn)行檢測(cè)器”按鈕。 提示會(huì)通知您有關(guān)運(yùn)行數(shù)據(jù)所需的處理配額數(shù)量。 由于我們使用的是免費(fèi)配額，因此請(qǐng)按開(kāi)始檢測(cè)。

Prompt for the Processing quota for the detector提示檢測(cè)器的處理配額

Viewing the results shows us the output of the detector

查看結(jié)果可向我們顯示檢測(cè)器的輸出

Detector output from unseen images來(lái)自看不見(jiàn)圖像的檢測(cè)器輸出

We can also use this powerful platform, for other object detection activities, like solar-arrays, ships, trash, military vehicles etc.!

我們還可以使用這個(gè)功能強(qiáng)大的平臺(tái)進(jìn)行其他物體檢測(cè)活動(dòng)，例如太陽(yáng)能電池板，輪船，垃圾箱，軍用車(chē)輛等！

Please check Picterra’s blog for their recent success stories:

請(qǐng)查看Picterra的博客，了解他們最近的成功案例：

• Object detection stories

  物體檢測(cè)故事

• Bring AI-powered object detection to ArcGIS

  將AI驅(qū)動(dòng)的對(duì)象檢測(cè)帶到ArcGIS

Note on image licensees:

關(guān)于圖片被許可人的注意事項(xiàng)：

Unless mentioned otherwise, the author owns the licenses to the images.

除非另有說(shuō)明，否則作者擁有這些圖像的許可證。

Indian mosiac (NO2 levels) was created by the author using Sentinel-Hub Platform

作者使用Sentinel-Hub平臺(tái)創(chuàng)建了印度洋霜(NO2含量)

Time series plots for No2 levels in Germany & Italy were created by the author using code specified above

作者使用上面指定的代碼創(chuàng)建了德國(guó)和意大利2級(jí)水平的時(shí)間序列圖

Screenshots are from Picterra’s geo-spatial platform

屏幕截圖來(lái)自Picterra的地理空間平臺(tái)

翻譯自: https://towardsdatascience.com/how-i-won-sentinel-hub-covid-19-custom-script-hackathon-be882ed05186

總結(jié)

以上是生活随笔為你收集整理的如何使用Picterra的地理空间平台分析卫星图像的全部?jī)?nèi)容，希望文章能夠幫你解決所遇到的問(wèn)題。

如果覺(jué)得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。