【学生实验】用人工智能防治森林火灾

暑假 AI 实践课堂上，我们一起带着满满的好奇心，了解全球气候变暖对地球带来的灾难，如冰川融化、海啸、森林火灾、干旱、城市高温，惊讶于40年间消失了一半的盐湖，一再对比着冰川不断内退的边缘，感叹新加坡是如何在楼宇间巧用水流和绿植给城市降温，好奇人工智能是如何预测一个个海啸的完整行程…

更过瘾的是学生自己反复训练 AI 模型来更精准识别各种灾难的发生，了解科学家和工程师们是如何使用卫星遥感、人工智能和无人机等工具来及时预测和应对这些灾难的。

在一次次突破知识边界的过程中，刚刚10岁的学生 Annie 也完成了自己与 AI 的第一次互动研究和第一篇实验报告。完整发表如下：

Wildfire and AI

By: Annie（陈千寻）

I would like to choose the topic of Wildfires for my research, and the reasons are:

With more and more forests burning, the oxygen    generated by forests decreases, which is not good for human life.

Wildfires can burn people’s faces, damage their appearance, and cause them to suffer from bullying. I don’t want people to be bullied. Humans could become extinct if we lose oxygen every day.

I carried out AI training on Google Teachable Machine by following the steps below:

Step 1: Go to Google Teachable Machine;

Step 2: Select image option;

Step 3: Select 2 different types of images, one for wildfire, and one for forests;

Step 4: Put wildfire ones into class 1, and put the normal forest images into class 2;

Step 5: Press start AI model training, and don’t switch to any other tabs;

Step 6: When the model is ready, you can put an image to test if the model can identify if it is a forest with wildfire or not, by percentage.

Step 7: If the result is not accurate, you can add more images for Class 1 and Class 2, as you may have mixed the Class 1 and Class 2 pictures previously and retrain the model again;

Step 8: By retraining the model, it becomes more and more accurate.

We also carried out our research by using satellite monitors on wildfires. This is how we did our research:

Step 1: Search on NASA Wildfire Map;

Step 2: Zoom out the map so we can see the wildfire worldwide by looking at red dots which indicate the wildfire;

We found that South Africa and South America have the most dots gathered.

We also found there are many red dots (wildfires) in Middle East.

We continue raising our questions that why these countries or regions have so many wildfire, and the main reasons we worked out are:

1) These countries are the hottest areas worldwide;

2) Human accidence or war could cause the wildfire.

My conclusion is:

I think this research is very important as a lot of people get killed or injured by the wildfire, so we can use our findings to save people’s lives. 

Drone can take pictures and then AI can helps to widen rectify the wildfire instantly so people can act asap and accurately. Without AI and satellite, it is going to be very hard to identify and predict the wildfire.

End of Project.