AI garbage just around the corner?

640?wx_fmt=gif

640?wx_fmt=jpeg

Author | Tibetan fox

This article is licensed public switched from No. brain diode (ID: unity007)


Shanghai since the beginning of the implementation of waste, to be crazy on Haining has become the majority of users happy fountain, he contributed a lot of scripts and expression package. It is said that some people drive to Hangzhou and Suzhou trash, there are many really do not understand the classification of the field "study slag" could not carry direct pressure back home ......


640?wx_fmt=png


Shanghai people whom headache fast time, before the second still think "ameonna no melon" busy "Ha ha ha ha ha" sand sculpture users will soon be a reality thunderstruck:


640?wx_fmt=png


Resistance is impossible to resist. People inherent in a death, or died in Shanghai's garbage, trash or other 45 died in the city. In short, no one even think about littering!


I now eat a proactive thing, to be subconsciously thinking about it in front of it - is exactly what you refuse? However, benefit of the doubt, 9012, and garbage so difficult thing, AI can not do it to look ah?



640?wx_fmt=png

You have a mature AI, and the learn to classified



Garbage is really a very wonderful topic, many kinds of harsh Japanese garbage classification policy will have 34 kinds, which are clearly defined throw garbage every day, missed time may also left at home.


And the industrial chain is very long and the subsequent collection, transportation, distribution and disposal are very difficult to see. So for speaking region just started to produce "I divided the class but the garbage truck is not the car side pot to go," the question is simply too normal.


In short, any link cut off, it will directly affect the ultimate effect of garbage classification.


Fortunately, the rise of AI technology, has been able to contribute for this global project.


For now, AI can provide assistance in the whole industry chain:


1. The distal end (end resident) Intelligent Detection


"Crazy" garbage difficulty of the general public, mainly lies in the recognition of different material characteristics and be categorized, which involves a relatively high technical threshold. Spread on the network, a lobster needs are divided into five parts delivery, to throw away the mouse must dismemberment, pearl milk tea drink half how to throw ...... not to mention ordinary uncle aunt, and highly educated students are afraid need some ignorant force.


The most appropriate way of course is to not eat (sentence crossed out) to deliver intelligent computer vision through it.


For example, the district has been installed on the smart garbage station, just will automatically recognize the type of residents was delivered before the trash can scan it, and suggesting specific classification. If it can be bought recyclable garbage, trash after delivery to the appropriate cash redemption will automatically hit the mobile phone account in the residents, can be said to be very suitable for the lazy.


当然,没有此类智能垃圾桶的小区,也有带有AI识别的功能的手机App助攻,比如支付宝最近推出的“垃圾分类助手”,就成了救上海人民于水火的神器。


640?wx_fmt=png


2.终端(回收者)自动化


不得不说上海不愧是城市化最高的地区之一,广大群众们吐槽归吐槽,但也都是极尽可能地配合垃圾分类政策,努力程度堪比高考。不过,在家分的再好,如果垃圾车全部都混为一体,或者不考虑小区的实际量级,那也会带来不少的麻烦,让大家做无用功的同时,也影响政策的公信力。


因此,提高回收环节的清理效率和分拣水平,就变得至关重要了,而这正是AI所擅长的。


举个遥远的例子,在硅谷,创业公司Compology 就给小区的垃圾箱配备了智能传感器。这些传感器每天会多次拍摄垃圾桶内部的高分辨率照片,并发送图像到云端。这样,垃圾清理公司就能够及时监控信息,优化卡车清运垃圾的路线或时间表,快捷高效地拾取垃圾,从而保证了不同规模小区的清理效率。


640?wx_fmt=png


而在运载过程中,垃圾分类后也导致清运车增加,从2月20日起,上海全市就配置及涂装湿垃圾车982辆、干垃圾车3135辆、有害垃圾车49辆以及可回收物回收车32辆。显然,分类的细化也会导致司机出现人手不足的情况,而自动驾驶则有望解决这一问题。


今年五月,沃尔沃公司宣布与瑞典的Renova公司联手,开始测试自动驾驶垃圾车。除了和普通无人车一样配置激光定位器、雷达、摄像头、红外摄像头等传感系统之外,这种卡车还能够按照设置好的路线,沿途收集垃圾。所以,驾驶员只需要走两步,专心收集垃圾,不需要每次都返回驾驶室,开着车再前往下一个垃圾桶,大大减少了停车次数。


640?wx_fmt=png


同时,垃圾回收汽车还能够起到终端网络的监测作用。


我们以上海的垃圾收运为例,每辆垃圾清运车行走到了哪里,在哪个小区运了哪些类型的生活垃圾,装进了哪个集装箱,运到哪里处置,这些实时数据都会上传到“城市的垃圾大脑”(真的没有骂人的意思TAT),然后城市环卫系统和再生资源系统会根据前端的数据进行分析,从而对垃圾清运、设施布局等城市行为作出更好的规划。


3.后端(处理厂)智能化


好了,经过人和机器的努力,咱们的垃圾们终于来到了处理厂,可以进行劳动改造了。


这里的问题也是最多的。


首先,再严丝合缝的前中端把控,也有可能造成漏网之鱼,比如将有害垃圾丢进了干垃圾里,这时候就需要识别出是哪个小区出了问题需要强化分类教育,同时,处理厂还要进行二次分拣。


但是,回收垃圾带给人类员工的伤害也是巨大的。传统垃圾分拣的工作是由人类来完成的,肮脏、枯燥,而且危险,常常会接触到有害物品,比如针管、碎玻璃等等,也被称为美国最危险的职业之一。


640?wx_fmt=png

(传统的人工垃圾分类处理流水线)


而处理工厂的智能自动化, 一旦能够普及应用,就可以让这些分拣工人离开那些危险的岗位了。


前不久,北美纸箱包装委员会就与阿尔卑斯废物循环利用,以及AMP机器人这两家公司合作,在工厂中安装了AMP公司的Cortex分类机器人。


这种机器人配备了像蜘蛛一样的机械臂,利用摄像机向云端大脑传递影像信息,机器学习算法识别出传送带上的废物,机械臂就会对其进行分拣。


目前,机器人能够达到高达98%的分类准确度,每天工作大约16小时,每分钟可以做出60次分拣动作,远高于人类每分钟40次的平均值。


640?wx_fmt=png


同样这么做得还有芬兰公司ZenRobotics机器人,在美国Recon废物服务公司中,安装了人工智能回收系统Heavy Picker,可以抬起60磅重的物体,能够整理建筑垃圾,将其分类成金属、木头、石头等,然后投入循环利用。目前,苏州绿和公司也引入了该技术。


能够直接解放人类分拣员,减少分拣环节的健康风险,并有效提升了分拣效率。


640?wx_fmt=png


看到这里,相信很多人已经可以感觉到,AI在垃圾领域的应用有哪些值得注意的特点了吧。


简单来说,一是依靠成熟的感知技术,比如传感器、计算机视觉等等,让每个环节流通的垃圾和行为都能被数据化。而要让识别的准确率足够高,也需要进行一定的数据积累与训练。换句话说,AI系统的引入宜早不宜迟。


二是云+端+边算力的综合保障。我们发现,垃圾分类所涉及的环节对实时动态数据的监测和处理要求非常高,无论是在垃圾倾倒时的实时甄别,还是车辆行进路线的合理控制,这个过程都需要基础算力的支持,因此,边缘算力、终端芯片、云端处理的综合联动才能成就这项庞大的城市工程。未来随着5G网络的普及,即时的数据观察会让AI的效能变得更强。


听起来很美,那么,AI的落地有没有什么限制条件呢?答案几乎是肯定的!



640?wx_fmt=png

 万事俱备,只欠……



AI的“垃圾分类”探索说到这里,相信很多人已经感觉到了好像哪里不对的样子。为了人类生存环境变得更好,AI似乎在透支当下一部分人群的生活状态。


最为直接的影响是,随着智能机器人的引入和垃圾处理场的自动化改造,会有很多从事驾驶和分类工作的工人失业。


的确,他们的工作条件称不上好,但也是一份能够养家糊口的谋生之道。让他们转型去做那些AI提供的新工作岗位,比如数据分析师、操控高科技卡车和设备的机械师,这可能吗?


未来垃圾清理产业的人员素质必然会大量提升,但机会未必真的会属于那些被机器淘汰的一线工人。届时大量的底层劳动人口去往何处,恐怕是一个棘手的问题,因此,垃圾产业智能化的步子应该不会迈的太快。至少,在很长一段时间内,还是会由人类大爷阿姨来为你答疑解惑,而不是AI。


640?wx_fmt=png


另一方面则是部署成本的问题。


关注我们的小伙伴可能常常听到智慧城市、车路协同之类的前沿技术名词。但目前很多综合方案都还在封闭道路上测试,或是刚刚开始终端改造。


而垃圾分类的收集终端密集,数据维度多样,有着较大的自由度和模糊地带,层出不穷的新型垃圾也在挑战着传统的分类体系,这就导致现阶段想要依靠AI实现精准判断和运维决策,几乎是一件不可能的事。而在一个400万人口的中等城市,建设智能收集终端+智慧平台+智能检测线的一次性投资,初步估算约15亿元左右,这些都是要城市财政来买单的,恐怕只有少数超级城市能够逐步启动。而其他二三线垃圾分类试点城市,恐怕只能多多学习下分类手册,打打扔垃圾小游戏了。


AI garbage future perhaps a long way off, but the future is still worth the wait. Everyone must grow with the times, from this perspective, we can be considered a witness to history now. Let us together and AI, to the years to civilization, not to civilization to years.


640?wx_fmt=jpeg

 Thermal paper  recommended 

Zhang talk about 5G harmful theory; China Unicom response "closed 2G, 3G networks"; PHP 7.3.7 released | Geeks headlines

engineer annual salary of 980,000 machine learning, AI but employment growth is falling!

Why the "big company ills of technology simply can not save"?

Baidu cooperation with Huawei heavy! Robin Li: Baidu's technology is faith

☞ because of the TA, to get industry application development, not afraid not afraid of you!

☞ God except V, 17 Ethernet General Assembly Square lecturer's speech in essence here!

☞ 2019 years of technical papers container inventory (b): Listen to talk into the container technology Tencent cloud maturity | hardcore programmers evaluation

☞50 line of Python code, obtain public the full article number

☞ without writing a single line of code, you can also be Fun Kaggle contest?

☞ Ma once idol, Ali finally left 140 billion lost light!

640?wx_fmt=png Your point of each "look", I seriously as a favorite

Guess you like

Origin blog.csdn.net/csdnnews/article/details/94682212