Clive Cookson in London
Report a mispronounced word
It is dry season in a Kenyan national park. A small group of poachers walks along a dried-up riverbed, aiming to kill a black rhino and remove its horns, which could fetch as much as $100,000 on the Asian black market.
The men are concealed by undergrowth on the riverbanks but seen by a poaching alarm system developed by the Zoological Society of London. Their guns and knives trigger the Instant Detect system’s hidden metal detector, which activates a camera camouflaged in a bush. The images travel by radio to a base station and then via a communications satellite to the park headquarters, alerting the authorities in time to dispatch rangers and catch the gang.
Similar scenarios will soon be playing out in reserves and parks around the globe, as conservation bodies adopt a high-tech approach in their battle to protect animals. Some of these groups were slow to see the potential for new monitoring tools, but with the aid of groups such as Google they are now embracing the devices as a way to tackle poaching.
WWF estimates the illegal wildlife trade is worth about $20bn a year and has contributed to a catastrophic decline in some species. According to the Living Planet Index maintained by WWF and ZSL, the FT’s Seasonal Appeal partner for 2019-20, 60 per cent fewer vertebrate animals (mammals, bird, reptiles, amphibians and fish) live wild now than 50 years ago, with the steepest drops in the tropics.
Although huge falls in populations of some well-known species such as tigers, elephants and black rhinos have been halted and even reversed through intensive conservation efforts, poachers are still killing them — while numbers of other animals, including pangolins and many monkeys, are declining fast.
There are many drivers behind the loss of biodiversity, from overfishing and climate change to urbanisation and local pollution. But for some species illegal trapping and killing is the biggest factor in their decline, says Andrew Terry, head of conservation at ZSL.
“We have a particular focus on tackling the international wildlife trade but that is embedded in our broader conservation efforts,” says Mr Terry.
Zoologists have used camera traps to photograph passing animals for decades but until recently these had no wireless connection, so their operators had to physically visit each one to remove its film and later its electronic SD card, which was often full of useless images of moving branches or other wildlife that had triggered the trap.
“The odd thing is that conservationists were slow to take up technology,” says Eric Dinerstein, director of wildlife and biodiversity at Resolve, a conservation charity based in Washington DC. “We jumped in around six years ago because we saw an opportunity to make a difference with a camera trap with intelligence and connectivity.”
Other conservation bodies including ZSL began to develop detection technology at about the same time, working with tech companies that see wildlife protection as a showcase for their expertise. Systems such as ZSL’s Instant Detect and Resolve’s TrailGuard are in the final stages of testing and will be soon ready for deployment in the field.
“Conservation organisations don’t generally have the resources to recruit and employ expensive software engineers and developers, so we depend on collaboration with the tech industry,” says Sam Seccombe, Instant Detect project manager. ZSL’s partners include Google and Iridium, the satellite communications operator, while Resolve is working with Intel, Microsoft and Inmarsat, another satellite company.
Google’s AutoML system, which enables people with limited expertise to develop artificial intelligence for specific purposes such as image recognition, is being deployed in Instant Detect, making it possible to recognise people or animals instantly from camera trap pictures.
“A successful business relies on being able to collect, analyse and interpret data rapidly to make the best business decisions,” Mr Seccombe says. “The same is true for conservationists using camera trap data. By increasing the speed of image analysis, conservation impact can be made more quickly and be more effective.”
Remote wildlife parks have little or no mobile phone coverage, so Instant Detect uses its own radio transmitters to send images to a buried base station and then on by satellite to headquarters. ZSL tested the first version of the system by monitoring Antarctic penguins, Canadian bears, Australian night parrots and Kenyan elephants and rhinos. But it suffered from transmission problems, particularly in dense foliage.
The team has developed a more robust and reliable second version, Instant Detect 2.0, which has had successful preliminary tests in Africa and will undergo more extensive trials in the new year in Thailand’s Western Forest Complex and elsewhere before operational deployment.
Camera quality was also an issue. Nothing on the market met ZSL’s specifications so it developed its own 5-megapixel Instant Detect camera with a wide range of focal lengths, triggered either by an inbuilt infrared sensor that detects heat and motion of a passing animal, or by an external metal detector for poachers. “It seems ironic that most trail cameras being used by conservationists have been designed for the deer hunting market,” says Mr Seccombe.
Although the camera has a powerful computer chip that could run an automatic image processing system on captured pictures to decide whether they are worth transmitting, this feature will not be used initially, so as not to overload the system. Instead, image processing will take place in the cloud after the images have been transmitted.
Another development in the near future will be the integration of acoustic sensors, triggered by sounds such as a gunshot, chainsaw, engine or animal call. ZSL is developing a machine-learning algorithm to detect shots in collaboration with Google.
Resolve’s TrailGuard, which incorporates Intel vision-processing chips in its cameras, does carry out AI image analysis locally, so that only pictures of human intruders are transmitted — extending battery life and cutting transmission costs. The first version of TrailGuard, operating in the Grumeti reserve in Tanzania last year, detected more than 50 intruders and enabled rangers to make 30 arrests from 20 different poaching gangs and seize 1,000kg of illegal bushmeat.
Mr Dinerstein says Resolve is manufacturing 1,000 updated TrailGuard units, 300 in California and 700 in China, for installation in parks in Africa and elsewhere. The US foundation proposes to protect 100 wildlife parks and reserves over the next two years, by placing TrailGuards on the 10 trails used most actively by poachers in each place. Once satellite modems have been installed, the number of cameras can be increased to as many as 100 per park.
Installation would cost a park an estimated $17,000 in the first year and slightly more in the second year, with future operating expenses for data transmission at about $200 a year — much less than alternative protection measures such as flying drones to spot poachers or employing additional rangers.
Anthony Dancer, who manages ZSL’s tech programme, warns that new technology cannot stop illegal killing on its own. “Most protected sites around the world are terribly underfunded,” he says. “Even if we make the technology available, many places will not have enough resources to manage the technology or enough rangers for a large increase in enforcement.”
Besides poaching for meat, horns, teeth, scales, fur and other valuable products, people also kill animals to stop them raiding crops or livestock. Resolve plans to tackle this growing conservation problem by adapting its TrailGuard hardware to identify animals rather than people, for a project called VillageGuard.
Camera traps, installed along trails used by large animals that eat or trample crops or attack livestock, will automatically detect intruders. The first five targets are elephants and lions in Africa, snow leopards and wolves in Nepal, and grizzly bears in the US. Attached speakers will then frighten away the unwanted animals with alarming sounds such as human shouting.
Beyond the detection of threats to wildlife from poachers or angry villagers, conservation bodies are enlisting information technology to track elusive animals. They analyse the rapidly increasing volume of images emerging from camera traps installed around the world. ZSL uses both machine learning and human volunteers for this task.
Several projects are under way to identify animals through AI. The largest collaborative programme, called Wildlife Insights, sits in Google Cloud and combines the company’s machine learning expertise with a group of conservation groups including ZSL. It has been trained to recognise 614 different species with 8.7m images supplied by member organisations — and expects to expand rapidly as conservationists feed in more data. Initial accuracy ranges from 80 per cent to 98 per cent, depending on the quality of the image and the distinctiveness of the species.
“Wildlife Insights is essentially a massive open source system that will enable people around the world to manage and analyse biodiversity data,” says Mr Dancer.
While artificial intelligence becomes an ever more powerful tool, humans will always play an essential role in wildlife identification — including amateur as well as expert zoologists. Instant Wild is ZSL’s free citizen science app that anyone with a smartphone can use to identify animals in camera images; it has been downloaded 130,000 times. Crowdsourcing analysis of this sort is useful for educating and involving the public, as well as directly assisting with species identification.
“You don’t need to be an expert. You just give your best guess,” says ZSL project manager Kate Moses. “The result only goes through to the project scientist when 10 people have given the same identification.”
Technology is also helping the people on the front line of the battle to protect wildlife: the 300,000-400,000 rangers and wardens who work in the world’s parks and reserves, according to the International Ranger Federation. A system called Smart (for Spatial Monitoring And Reporting Tool), developed by ZSL and other conservation bodies, enables rangers to collect and sort out data on their mobile devices about the locations of animals and humans, including illegal intruders, in order to deploy scarce staff as efficiently as possible.
Smart is already being used in 900 protected areas around the world. AI software developed by Harvard computer science professor Milind Tambe will be integrated into the system next year. This predicts poachers’ behaviour, so that patrols can be directed to likely hotspots of illegal activity.
With animal populations under enormous pressure, technology has huge potential for enabling conservation groups to deploy their resources more efficiently in the battle against poaching and the wider illegal wildlife trade.
“We urgently need to innovate, and to create new partnerships with industry, governments and academia, to develop new solutions,” says Mr Dancer of ZSL. “This is where technology and tech partnerships have the potential to be transformational — by enabling us to target our resources more efficiently and more effectively, and to scale our impact.”
The primary aim of conservationists is to stop poachers killing animals but, when they fail, advances in forensic science can help to catch criminals in the illegal wildlife trade.
Researchers are enhancing fingerprinting technology to improve the chances of obtaining clear prints from people who have handled animal parts.
Working with colleagues at the University of Portsmouth, ZSL scientists are using “gel lifters” — small sheets coated with sticky gelatin — to remove fingerprints from pangolin scales and other unpromising materials such as snake skins. The prints are then read with specialist imaging machines.
Another collaboration, involving City of London police and King’s College London, has developed a new magnetic powder that enables investigators to recover human fingerprints from elephant tusks with much better definition than conventional methods. It can recover prints up to four weeks old, giving more time to gather evidence on criminals who have handled ivory seized by police or customs officers.
Conservationists are also using new genetic analysis in two ways to investigate wildlife crime. First, if poachers leave tiny traces of their DNA on ivory, horn or other material, it may be possible to track them down through a genetic fingerprint.
Second, animal DNA extracted from illegally traded materials can be used to trace its geographical origin. This prospect may be particularly applicable to smuggled ivory, as scientists build up a database showing the genetic differences between elephant populations in different parts of Africa.
Scientists at Liverpool John Moores University have used ultrasensitive DNA probes to help spot illegal animal material within large cargo consignments at borders, ports and customs posts. In a proof of concept study published this month they rapidly identified tiny quantities of tiger, rhinoceros and pangolin DNA.
Please help us support ZSL’s urgent work by making a donation to the FT’s Seasonal Appeal. Click here to donate now.
If you are a UK resident and you donate before December 31, the amount you give will be matched by the UK government — up to £2m. This fund-matching will be used to help communities in Nepal and Kenya build sustainable livelihoods, escape poverty and protect their wildlife.
Read more about our Seasonal Appeal partner ZSL: ft.com/zsl-facts
Get alerts on FT Seasonal Appeal when a new story is published
伦敦的 Clive Cookson
随着环保机构在保护动物的斗争中采用高科技手段，类似的情况很快将在全球的保护区和公园上演。这些组织中的一些人对新的监控工具的潜力认识缓慢，但在谷歌( Google )等组织的帮助下，他们现在开始将这些设备作为对付偷猎的一种方式。
世界野生动物基金会( WWF )估计，非法野生动物贸易每年价值约200亿美元，导致一些物种的数量大幅下降。根据世界自然基金会( WWF )和英国《金融时报》2019-20年度《季节性呼吁》( Seasonal Appeal )合作伙伴 ZSL 维护的“活行星指数”( Living Planet Index )，如今在50年前，生活在野外的脊椎动物（哺乳动物、鸟类、爬行动物、两栖动物和鱼类）减少了60%，热带地区的降幅最大。
生物多样性丧失的背后有许多驱动因素，从过度捕捞和气候变化到城市化和地方污染。但是对于一些物种来说，非法捕杀和杀戮是它们减少的最大因素， ZSL 的保护主管 AndrewTerry 说。
动物学家们用相机陷阱拍摄经过的动物已经有几十年了，但直到最近，这些动物还没有无线连接，所以他们的经营者不得不亲自去看他们的电影，后来他们的电子 SD 卡，这往往是充满无用的图像移动分支或其他野生动物触发了陷阱。
“奇怪的是，自然资源保护主义者在技术上进展缓慢，”总部位于华盛顿特区的保护慈善机构 Resolve 野生动物和生物多样性主管埃里克·迪恩斯坦( Eric Dinerstein )说。“大约6年前，我们跳了进来，因为我们看到了一个利用智能和连接来改变摄像机陷阱的机会。”
包括 ZSL 在内的其他保护机构大约在同一时间开始开发检测技术，与那些将野生动物保护视为其专业知识展示的科技公司合作。ZSL 的即时检测和解决方案的 TrailGuard 等系统正处于测试的最后阶段，并将很快在现场部署。
“保护组织通常没有资源招聘和聘用昂贵的软件工程师和开发人员，因此我们依赖于与科技行业的合作，” Instant Detect 项目经理萨姆•塞科姆( Sam Seccombe )表示。ZSL 的合作伙伴包括谷歌( Google )和卫星通信运营商铱星( Iridium )，而 Resolve 正与英特尔( Intel )、微软( Microsoft )和另一家卫星公司 Inmarsat 合作。
谷歌( Google )的 AutoML 系统正被部署在即时检测( Instant Detect )中，使人们能够从相机陷阱图像中即时识别人或动物。该系统使专门知识有限的人能够为图像识别等特定用途开发人工智能。
照相机质量也是一个问题。市场上没有任何东西符合 ZSL 的规格，因此， ZSL 开发了自己的500万像素的即时检测摄像头，具有广泛的焦距，可以通过内置的红外传感器检测过路动物的热量和运动，也可以通过外部的金属探测器检测偷猎者。“似乎具有讽刺意味的是，自然资源保护主义者使用的大多数履带摄像机都是为猎鹿市场设计的，”塞科姆表示。
在不久的将来，另一个发展将是声音传感器的集成，由枪声、链锯、发动机或动物叫声等声音触发。ZSL 正在与 Google 合作开发一种机器学习算法来检测照片。
Resolve 的 TrailGuard 将英特尔的视觉处理芯片集成到相机中，它在当地进行人工智能图像分析，因此只有人类入侵者的照片才能传输——延长电池寿命并降低传输成本。第一个版本的 TrailGuard 去年在坦桑尼亚 Grumei 保护区作业，发现50多名入侵者，并使游骑兵从20个不同的偷猎团伙中逮捕30人，并没收1000公斤非法丛林肉。
丁尔斯坦表示， Resolve 正在生产1000个更新的 TrailGuard 单元，其中300个在加利福尼亚，700个在中国，用于在非洲和其他地方的公园安装。美国基金会提议，在未来两年内保护100个野生动物公园和保护区，在每个地方的偷猎者最积极使用的10条小径上放置 TrailGuards 。一旦安装了卫星调制解调器，每个公园的照相机可以增加到100架。
管理 ZSL 技术项目的安东尼•舞者( Anthony Danger )警告称，新技术不能单独阻止非法杀戮。“世界上大多数受保护的网站资金严重不足，”他说。“即使我们提供这项技术，许多地方也没有足够的资源来管理这项技术，也没有足够的护林员来大幅提高执法力度。”这是一个很好的例子
除了偷猎肉类、角、牙齿、鳞片、毛皮和其他有价值的产品外，人们还杀死动物来阻止它们袭击庄稼或牲畜。解决这个日益增长的保护问题的计划，通过调整 TrailGuard 硬件来识别动物而不是人，为一个叫 VillageGuard 的项目。
一些通过人工智能识别动物的项目正在进行中。最大的合作项目名为“野生动物观察”( Wildlife Insights )，位于谷歌云( Google Cloud )，将该公司的机器学习专长与包括 ZSL 在内的一组保护组织结合起来。该公司接受了识别614种不同物种的培训，成员组织提供了870万张图像，预计随着环保人士提供更多数据，该公司将迅速扩张。最初的准确度从80%到98%不等，这取决于图像的质量和物种的独特性。
“野生动物观察( Wildlife Insights )基本上是一个庞大的开源系统，将使世界各地的人们能够管理和分析生物多样性数据，” Danger 表示。
虽然人工智能成为越来越强大的工具，但人类在野生动物识别中始终扮演着至关重要的角色，包括业余和专家动物学家。即时野生（ Instant Wild ）是 ZSL 的免费公民科学应用程序，任何拥有智能手机的人都可以用它来识别照相机图像中的动物；它已被下载了13万次。这种类型的众包分析对于教育和公众参与以及直接协助物种识别是有用的。
“你不需要成为专家。ZSL 项目经理 KateMoses 说。“当10人给出了相同的识别结果时，结果只会传给项目科学家。”
国际游骑兵联盟( International Ranger Federation )的数据显示，科技也在帮助保护野生动物的前线人员：在世界公园和保护区工作的30万至40万游骑兵和看守员。一个名为 Smart （空间监测和报告工具）的系统，由 ZSL 和其他保护机构开发，使游荡者能够收集和整理移动设备上关于动物和人类位置的数据，包括非法入侵者，以便尽可能高效地部署稀缺的工作人员。
智能已经在世界各地900个保护区使用。哈佛大学计算机科学教授 MilindTambe 开发的人工智能软件将于明年集成到该系统中。这预示着偷猎者的行为，因此巡逻可能会被导向非法活动的热点地区。
“我们迫切需要创新，并与行业、政府和学术界建立新的合作伙伴关系，以开发新的解决方案，” ZSL 的 Danger 表示。“这是技术和技术伙伴关系具有变革潜力的地方——通过使我们能够更有效和更有效地瞄准我们的资源，并扩大我们的影响。”
与朴茨茅斯大学的同事们合作， ZSL 的科学家们正在使用“凝胶升降机”——涂有粘性明胶的小薄片——来从穿山甲的鳞片和蛇皮等其他未被发现的材料中去除指纹。然后用专门的成像机读取打印.
伦敦金融城( City of London )警方和伦敦国王学院( King ’ s College London )参与的另一项合作，开发出了一种新的磁粉，使调查人员能够用比传统方法更好的定义从大象象牙中提取人类指纹。它可以恢复四个星期前的打印，给了更多的时间收集罪犯谁处理了由警察或海关官员没收的象牙。
自然资源保护主义者也在用两种方法来研究野生动物犯罪.首先，如果偷猎者在象牙、角质或其他材料上留下微小的 DNA 痕迹，就有可能通过基因指纹追踪它们。
其次，从非法交易材料中提取的动物 DNA 可用于追踪其地理来源。这一前景可能特别适用于走私象牙，因为科学家建立了一个数据库，显示非洲不同地区大象种群之间的遗传差异。
利物浦约翰·摩尔斯大学的科学家利用超灵敏的 DNA 探针帮助在边境、港口和海关的大型货物托运中发现非法动物材料。在本月发表的一项概念研究的证据中，他们迅速发现了少量的老虎、犀牛和穿山甲 DNA 。
请通过向英国《金融时报》的季节性呼吁捐款，帮助我们支持 ZSL 的紧急工作。单击此处立即捐款。
阅读更多关于我们的季节性呼吁合作伙伴 ZSL : ft 。com / zsl-事实
当一个新故事发布时，请获取 FT Seasonal Appeal 的提示