Entry Name:  "Visual Satellite Images Explorer"

VAST Challenge 2017

Mini-Challenge 3

Team Members:

Guozheng Li, Peking University, guozheng.li@pku.edu.cn     PRIMARY
Shuai Chen, Peking University, shuai.chen@pku.edu.cn

Qiusheng Li, Qihoo 360, qiushengli245@gmail.com

Zhibang jiang, Peking University, zhibang.jiang@gmail.com

Yuening Shi, Peking University, 1300012980@pku.edu.cn

Qiangqiang Liu, Peking University, qiangqiang.liu@pku.edu.cn

Xi Liu, Qihoo 360, sisiliu1204@gmail.com

Xiaoru Yuan, Peking University, xiaoru.yuan@gmail.com

Student Team: NO

Tools Used:

D3.js

Photoshop

OpenCV

Matlab

Approximately how many hours were spent working on this submission in total?

300hours

May we post your submission in the Visual Analytics Benchmark Repository after VAST Challenge 2017 is complete?

Yes

Video

http://vis.pku.edu.cn/~shuai.chen/mc3.mp4 

Questions

1 –Boonsong Lake resides within the preserve and has a length of about 3000 feet (see the Boonsong Lake image file). The image of Boonsong Lake and is an RGB image (not six channels as in the supplied satellite data). Using the Boonsong Lake image as your guide, analyze and report on the scale and orientation of the supplied six-channel satellite images.

Figure1: Satellite Image and Edge Detection Results.

In order to identify the Boonsong Lake in the image, we do the edge detection for the satellite image under the channel B3B2B1 in Aug. 24th, 2014 to get the edge detection result as shown in Figure 1.

Figure 2: Boonsong Lake in the satellite image.

By comparing the shape of the lake, we detected that Boonsong Lake is located in the southwestern of the satellite image, as shown in Figure 2.

As stated above in the question, Boonsong Lake has a length of 3000ft and is oriented north-south in the satellite image it has 35px totally in the north-south orientation, so we learned that the scale of the satellite image is 1px:86ft.

From the satellite images, there are roads around the Boonsong lake, which are straight and easy to get its orientation. The intersecting angle between the two same roads as shown in Figure 3 is 2.8 degrees through the formula below.

Then we could get the orientation of the satellite images, 2.8 degrees east of due north.

Figure 3: The comparison of the the orientation between the satellite Image and given image. 1 - road orientation in the provided image. 2 - road orientation in the satellite image. 3 – intersecting angle.

2 –Identify features you can discern in the Preserve area as captured in the imagery. Focus on image features that you are reasonably confident that you can identify (e.g., a town full of houses may be identified with a high confidence level). Please limit your answer to 6 images and 500 words.

Figure 4: The Overview of Collaborative Tagged System. 1. System overview. 2. Image tagged view. 3 Image comparison view. 4. Distribution comparison histogram. 5. Event list view.

The features in preserve are detected through tagging manually and collaboratively, and they are described from natural features and human activity perspectives respectively.

1. Natural Features

Natural features are stated from the overall and detailed topographic perspectives.

l Overall Features

The preserve is located in the Northern Hemisphere. It is consistent between icing phenomenon and Northern Hemisphere’s seasoning rule. As Figure 5(3), 5(4) under B1B5B6 channel show, the icing parts in red are observed on Dec 30th.

Secondly as shown in Figure 5(1), 5(2) under B5B4B3 channel, the shadow is located in the north of cloud itself, more northern in spring compared to summer.

The preserve appears an overall trend of high in east and low in west. Under channel B1B5B6 as shown in Figure 5-3, icing phenomenon, which is encoded in red, mainly occurred in the east.

Figure 5: The Overall Features of the Preserve. 1,2 with cloud and its shadow selected; 3,4 with ice detected, icing phenomenon is observed in December and none in June.

l Mountain: In Figure 6(1) under channel B4B3B2, the red regions are covered with plant, which are separated by roads. Also, in Figure 6(2) under the channel B5B4B2, the blue area is covered with ice in winter.

l Wetland: As the selected area in Figure 6(3) shows, the soil in dark means that it contains water, but in Figure 6(4) under the channel B1B5B6, which encodes ice in red, means that there is no ice, so we speculate that the region is wetland.

l Lake: The lake is in red under channel B1B5B6 in Figure 6(5), which means the lake is freezing in winter. In Figure 6(6), under channel B1B5B6, pure water appears in black, consistent with the selected area.

Figure 6: Detailed Natural Features. This figure shows the mountain, wetland, lake and rivers respectively.

2. Human Activity Fatures

l City: According to Figure 7(1), (2) under channel B5B4B2, the cities which are located in the northwest of the satellite image are encoded in purple.

l Road: Comparing with the rivers, the roads in the preserve are much more straight and keeps consistent color in different seasons. Roads are built between the detected cities and across the whole preserve.

Figure 7: Human Activity Features. 1 and 2 mark the cities under channel B5B4B2. 3 and 4 mark the roads in the preserve.

l Planting Area: Channel B3 is sensitive to different plant categories. In Figure 8, we mark out the region in the northwest. We found that plants in marked regions increased while there is an apparent decline outside the marked region, so we speculated that the marked region is the human planting area and the other is forest.

Figure 8: Planting Area. 1 and 2 show different tendency between planting area and other regions.

3 – There are most likely many features in the images that you cannot identify without additional information about the geography, human activity, and so on. Mitch is interested in changes that are occurring that may provide him with clues to the problems with the Pipit bird. Identify features that change over time in these images, using all channels of the images. Changes may be obvious or subtle, but try not to be distracted by easily explained phenomena like cloud cover. Please limit your answer to 6 images and 750 words.

The event detection is based on the identified features in the MC3-Question2. Comparing with the same features in different images, we could learn the changes in the preserve.

1. Overall Changes

The plant health and water content of the preserve was worse from 2014 to 2015, and then became better from 2015 to 2016.

From these result images (Figure 9(1), (2), (3)) of NDVI, the changes of plant health are displayed. Red regions in the images means the luxuriant plant. Selecting the similar date around September in 2014, 2015 and 2016, the plant health became worse firstly and then became flourish again.

NDWI is used to monitor changes related to water content. In the figure9(4), (5), (6), the red regions represent the water. From these figures, the changes of water are consistent with the plant health from 2014 to 2016.

Both water and vegetation became worse in 2015 then better in 2016, so we guess that the preserve become drought in 2015, and it influences its vegetation. In 2016 both of the precipitation and vegetation recovered again.

Figure 9: Plant Health and Water Content Changes. 1, 2, 3 show the plant health of the preserve under NDVI. 4, 5, 6 show the water content of the preserve under NDWI.

2. Detected Events

Flooding: From the figure below, in Nov 28th, 2014, there is lots of cloud during this period, so we guess that there might be continuous heavy rain these days and it caused the flooding in this area. As Figure 10 under the NDWI channel shows, the water content in Dec 30th, 2014, which is close to Nov 28th, 2014, is much more than Feb 15th, 2015 apparently and this phenomenon also validate the flooding events.

Figure10: Flooding Event. This figure shows the flood event in the figure B5B4B3 and NDWI channel.

Eutrophication: B4B3B2 shows the plant health in the preserve. From June 24th, 2015 to June 26th, 2016, the plants appeared in the marked lake. This phenomenon may be generated by the eutrophication of water bodies in the lake.

Making the exploration for the single channel image, B3, which is sensitive to the mineral deposition, and we found that the mineral content is much higher in June 24th, 2015 than June 26th, 2016, which also proved our eutrophication hypothesis, and we speculate that the reason is that the pesticide, fertilizer, and mineral in the soil flow into the lake with the flooding.

Figure11: Plant Health and Mineral Deposit. The lakes of the preserve are marked in the image, and 1, 2 are under channel B4B3B2, which shows the plant health in the preserve. 3, 4 are generated under channel B3.

Fire: We found that there was a fire between Aug 24th, 2014 and June, 24th, 2015, and the vegetation has been recover a little from June 24th, 2015 to June 26th, 2016.

B5B4B2 is used to show newly burned ground. In Figure 12, the purple regions represent the bare ground. Compared the selected regions (in the green circles) in the Figure 12(1), (2) and (3) we could find the bare ground appeared in 2015 and recover a little in 2016.

Figure 12: Fire Event. The fire occurred in the marked region and images are generated under the channel B5B4B2. The area in red is the newly burned ground.

Water Pollution: Under the channel B5B4B2, bared ground is encoded in red. In Dec 19th, 2016, we found that the color of selected district in the lakes is similar to bared ground. Also, the district is adjacent to the road, so we speculate that pollution is dumped into this lake, and the pollution contains some specific mineral.

Figure 13: The Water Pollution Event. The marked lake shows the regions of water pollution event and the images are generated under the channel B5B4B2.

Urbanization: Based on the detected city districts, and we could learn that this part is becoming larger and darker along the time. The distribution histogram also supports users’ exploration. From the Figure 13, we could see that the red channel and blue channel, which could be combined to get purple, is become more with the evolution of time, so we speculated that during these three years, there is a trend of urbanization.

Figure 14: Urbanization. The marked area shows the city regions. The distribution in the right shows the values distributions in different channel of the selected region.