本集內容

Shark-detecting drones 可探測鯊魚的無人機

學習要點

有關“Danger 危險”的詞匯

邊看邊答

How does the drone know if it's spotted a swimmer or a shark?

文字稿

A shadow in the sea or something more sinister?

From the beach it's hard to tell but from the skies the drone has a clearer view. It feeds into deep learning software which it's claimed has a 92% success rate in spotting sharks, as well as less threatening species. Every time it sees a dolphin, whale or a swimmer in distress it learns their shape. Launched along the beaches of New South Wales the drones can patrol for 40 minutes. Daniel was one of the first to train as a lifesaver pilot. 

Daniel Trollope, Drone pilot

I mean, I can't physically run out and grab a board and paddle out and save someone, but sitting on the beach we’ve got eyes in the sky and we're just another layer of protection, really. 

The drones don’t only observe - they can react too, dropping an inflatable device to help people in the sea. It won't replace the use of controversial shark nets and drum lines however – which some claim do more harm than good. The drones won't be on every beach every day, but they should give everyone a better understanding of what’s down below. 

僅僅是一個影子？或是一個更危險的生物？ 

站在沙灘上看去，很難辨認出海上的陰影是不是鯊魚，但空中無人機就能看到更清楚的圖像。無人機將拍攝到的影像傳輸給分析海洋的智能軟件，這個軟件號稱可以在探測包括鯊魚和一些沒那么危險的生物時達到92%的成功率。每當無人機看到遇險的海豚、鯨或游泳者時，便會記下一個新形狀。在澳大利亞新南威爾士州的海灘沿岸上首次推出了利用無人機執行海上巡邏探測工作，每次探測時間達40分鐘。丹尼爾是第一批接受操作無人機培訓的“救生飛行員”。

丹尼爾·特羅洛普 無人機操作者

“雖然我不能親自跑過去、拿上滑板，然后劃過去救人，但我們坐在海灘上通過“天上的眼”可以一睹海里的情況，從而我們也就可以給大家提供多一層的安全保障。”  

無人機不僅能檢測到海上的情況，還能做出及時的反應，比如扔下一個充氣漂浮物來幫助在海上的人們。然而，它不能代替尚存爭議的鯊魚網和鼓線的使用，有些人認為使用鯊魚網和鼓線的弊大于利。無人機上不會每天在每個海灘上巡邏，但它的使用應該能給人們提供一個更好地了解海底下情況的機會。

詞匯

sinister
有兇兆的，危險的 

threatening
威脅（性）的，危險的

in distress
遇險

do more harm than good
弊大于利

你知道嗎 ？

There are over 500 species of shark, ranging from the colossal 20m whale shark, which weighs around 30,000kg and is the world's biggest fish, to the tiny dwarf lantern shark that only grows to 10-20cm.

據統計，目前鯊魚的種類超過500種，從體重約3萬公斤的“世界上最大的魚”— 鯨鯊，到長度只有10到20厘米的“魚界的小矮人”燈籠鯊。

問題答案

It compares the shape against images and shapes stored in its deep learning software. It’s claimed to have a 92% success rate.