Smart City – How Does Computer Vision Benefit the General Public Without Invading Privacies? [智慧城市 – 電腦視覺如何在不影響個人私隱的前提下為大眾服務?]


09.07.2021

Weeks ago, a boar was in the limelight for invading an MTR station and traveling across the harbor. The incident is a piece of rare, fun news as long as the boar is unoffending. Otherwise, if the boar is furious, injuries and chaos might be caused and turn the funny matter into tragedy. Being overcrowded, wild animals may occasionally intrude into the city by accident. It is not possible to pay attention restlessly to the actions of the wild animals around the border dividing the city and the suburb because of the limitation of human resources. Then how about technology? How can we make great use of the advancing technologies to track down the wild animals before they intrude into the city, and prevent similar incidents from happening again?

Ever since the government published the “Hong Kong Smart City BluePrint”, the concept of Smart City has started to be understood by the general public. However, whenever the term is mentioned, most of the citizens may come up with the topic of smart lampposts or face recognition systems, which leads to arguments about the concerns of the invasion of privacy and over-monitoring by the government. In this day and age. It is inevitable to adopt the use of smart city technology to maintain the competitiveness of our city. With all the concerns and constraints, engineers are trying their best to strike a balance between privacy protection and the application of computer vision and other smart city technologies.

 

What is Smart City and Computer Vision?

Why are they related to the privacies of the general public?

 

To start with, let’s have a deeper understanding of the concept of “Smart City”. Smart City refers to using Internet of Things (IoT) technology, sensors, and other smart devices to gather different daily data of the city. Artificial Intelligence (AI) will then be applied to analyze, learn and apply these data and assist the managers (such as tunnel officers, building attendants, etc.) to make better decisions and manage the city with better efficiency and organization.

Using lenses and closed-circuit television to record and analyze data is one of the examples mentioned in the smart city blueprint, and this measurement is related to computer vision. Computer vision refers to teaching computers and AIs to process, categorize, and label the recorded images, and then detect and mark down the required information from the processed images, such as counting object amounts, marking facilities on a map, measuring population density, and so on.

Human face recognition is the most discussed topic recently and there is no gainsaying that it is a kind of computer vision technology. Yet, apart from recognizing human faces, computer vision could be used to detect other objects, such as public facilities, objects, accidents, and more. Applying computer vision technologies in the right places, the city, and the neighborhoods will be able to develop more securely and efficiently.

Examples of Computer Vision

Apart from common examples like the observation of road traffic or pedestrian density, computer vision technology is actually widely adopted in our daily lives.

Generating Data for Public Facilities

Computer vision may help the government to collect the data of public facilities. Take those on the pavements as an example. When the government wants to calculate the number of traffic lights or fire hoses and markdown their locations on the map for future use, related departments may take pictures on roads and streets with street view cars. However, if computer vision was not applied, these pictures are nothing but different pixel values. Extra human resources are required to sightsee and mark down the actual situation or data, the costs will increase and it would be more time-consuming. Once computer vision is applied, the artificial intelligence installed will automatically analyze the images, as well as locating different public facilities automatically, so as to assist the relevant departments to acquire demanded information more efficiently.

Fall Detection

As the name suggests, computer vision can be used to detect if pedestrians need any help. Fall detection can be applied in districts, estates, or buildings that are guarded by security. Use the clubhouses of some private properties as an example. The clubhouses are very large and separated by different floors. Although there are CCTVs monitoring everywhere, it is hard for the security guards to observe all the screens at once with only their eyes. They might not be able to react and provide assistance immediately. Computer Vision would be the remedy for this problem. Via real-time image analysis, the security guard will be alerted of the location when accidents happen, so they will be able to handle the accident promptly, to prevent the situation from getting worse.

Smart Mobility

How “smart” can traveling be? In fact, computer vision has also been applied to public transport. Sensors and lenses are installed around taxi stations or bus stops, to keep track of the queue length, the computer will then estimate the waiting time based on the historical data. The individuals may check out the real-time traffic information and make better plans for their journeys. The computer may even determine whether the waiting time is too long, and recommend travelers with alternative transportation options. Not only can the travelers save time for their journeys, but the stations will also be less overflowing, efficiencies of both sides are also enhanced.

Lenses and CCTVs are applied to the above examples to detect the movements of humans, so as to collect required data for different uses. It is rational for the general public to worry about privacy issues. Yet, the advanced technology has a great improvement in privacy protection. If human faces are captured by the lenses, they will be blurred by the AI automatically before the screenshots are analyzed. After the analysis process, only the data will be kept, the pictures will be deleted to ensure the privacy of involved pedestrians is protected.

Other common computer vision includes:

Traffic Observation

The Hong Kong Transport Department installed cameras beside the pavements and monitored the real-time traffic situations, such as recording vehicle flow amounts, car accidents, or traffic jams. The cameras focus on recording the profiles of vehicles and the overall usage of the roads, to calculate the related data. With this data, HKTD will be able to generate different information and deliver it to the road users via applications, websites, or radios, for them to plan a better route.

Fire Detection

Smoke detectors are installed on the ceilings to detect fire incidents, but when the alarms are triggered, there is already an accident. Smoke detectors cannot detect fire outdoors, nor can they prevent fire accidents, so how do we solve the problem with computer vision? Using CCTVs and infrared sensors, the recording focus of the lenses is the fire black spot and temperature of the environment. Smokes, flames, or abnormally high-temperature objects will be recognized by the computer, and the security guard will be alerted afterward. Hence, the security guards will be able to react immediately and contact the fire department to prevent fire accidents, so as to ensure the safety of the residents and assets.

Last but not least, let’s look back to the boar incident. If we are able to apply computer vision to track down the trace of the wild animals and send them back to the suburbs, it is believed that the chances for similar cases to occur will be lowered. Also, as mentioned, apart from improving the operational effectiveness of the city, computer vision may even ensure the safety of the general public.

Follow us on Social Media!

Facebook | LinkedIn | Twitter
______________________________

就在上兩週,一隻野豬走入港鐵站並登上列車成功「過海」,成為城中熱話。這件無疑是有趣且新奇的事,十年難得的一遇。幸而在這次事件中,走入港鐵的是沒有攻擊意圖的一隻小野豬,不然有可能造成混亂或受傷,趣聞變橫禍。城市人口稠密,野生動物不慎闖入鬧市的案例不時都會發生。始終人力有限,我們無法時時刻刻關注著所有效區與城市之間的邊界,只為監視著野生動物的一舉一動。那麼科技到底能否成為一個應對策略,在野生動物闖入前及時發現並牠們引導回郊區,避免同類事件再次發生?

自從政府在2017年12月公布了《香港智慧城市藍圖》後,智慧城市這個概念開始廣為人知。但每當提到智慧城市,相信大部分市民都會想起智能燈柱、人臉識別等議題,引起社會各界對監控和個人私隱安全關註的討論和熱議。然而,若我們想要確保城市能持續發展,在科技日新月異的時代保持競爭力,提升對智慧城市的應用是不可或缺的。既然我們無法避開城市趨向智慧發展,工程師正努力在應用電腦視覺科技發展智慧城市和保障廣大市民的個人私隱中取平衡。

甚麼是智慧城市和電腦視覺(Computer Vision)?
為何它會引起社會就個人私隱安全關註的熱議?

 

首先我們先來重新了解一下智慧城市這個概念吧。智慧城市是指應用物聯網(IoT)技術、感應器和其他智能裝置等,收集日常城市數據。並使用人工智能(AI)和電腦去分析、學習和應用這些城市數據,去幫助管理者(例如隧道交通督導員、物業管理者等)制定更好的決定,更有系統和有效率地管理這座城市。

至於在智慧城市概念中所提及的,使用鏡頭、閉路電視等裝置記錄和和分析城市數據,這些便涉及到電腦視覺(Computer Vision)的領域了。電腦視覺是指教導電腦和人工智能,對收集回來的影像進行圖象處理、分類和標籤,再從處理過的圖象中偵測和標記出所需的資訊,例如統計物品的數量、在地圖上標著公共設施、測量人口密度等。

當中引起爭議的便是人臉辨識。雖然我們不能否認它是電腦視覺技術中圖象辨別的一種,但除了辨識人臉這種圖象辨別方法外,電腦視覺技術也能應用在辦別其他事物。例子有公共設施、物品及意外偵測等,有助城市和社區更高效和安全地發展。

電腦視覺的例子

除了一些常見的例子如道路車流觀察、行人密度等,其實電腦視覺早己在不知不覺間應用在我們的日常生活中。

公共設施統計

電腦視覺亦能為政府統計公共設施的數據。以道路設施的統計為例:以往,若政府希望統計某一區的交通燈/消防喉的總數;及在地圖上標記它們的所在位置,存入資料庫以方便日後應用。雖然政府可以利用街景車將各街道及馬路的實際情況拍攝記錄下來,但在不應用電腦視覺分析影像的情況下,這些街景影像對電腦而然只是不同的像素值。政府可能需要額外派出人手實地勘察和紀錄下所需的數據,成本高昂而且十分費時。相反,若應用電腦視覺技術分析影像,人工智能可以自動對街景影象進行統計及分析,並自動找出不同的公共設施和它們相對位置,當局便能以更低成本及更高效率獲取需要的資訊。

跌倒偵測

顧名思義,這是一個運用閉路電視鏡頭和電腦視覺系統偵測是否有行人需要幫助的例子。這種技術可以應用在設有護衛員的大型社區、大樓或屋苑。以大型住宅項目的住客會所為例,由於這些設施的面積一般相對較大,而且分為不同樓層。即使有設有閉路電視系統及中央保安室集中顯示不同樓層和區域的畫面,護衛員亦很難以肉眼「24/7」全天候觀察到每一個畫面,為有需要的人提供即時協助。電腦視覺系統便能針對這一項以人手觀察所帶來的的不足之處,通過對多個實時影像進行分析,即時知會管理人員需要幫助的個案。以協助管理人員和護衛員及時伸出援手,提升處理意外的速度和效率,保障顧客和住客的安全。

智慧乘車

怎樣才算有「智慧」地乘車?其實智慧乘車概念也使用了電腦視覺這項技術,透過安裝在的士站/巴士站附近的鏡頭感應,記錄下當時正在等候汽車的人流。再電腦配合分析現時人龍和過往輪候時間數據,推測出實時的預計候車時間。用家可以透過手機應用程式,查看實時交通資訊,方便用家預算出發時間/決定合適的交通安排。電腦甚至能辨別出候車時間是否過長,推薦使用者到其他附近的的士站上車。除了能節省市民的候車時間,亦能為的士站分流,在兩方面提升整體效率。

以上三個例子都涉及到使用鏡頭或閉路電視功能去偵測人的行為,以收集合適的數據,用於不同用途。無可避免的是,使用鏡頭或閉路電視功能去偵察和分析影像的確會引起不少市民的擔憂,擔心其肖像或身形會被錄下和分析。但其實現時的電腦視覺技術在個人隱私方面已大有進步,即使鏡頭不小心將路人的肖像攝下,電腦 AI 也會先自動將人臉模糊化,才會開始分析影像。另外,這些分析用的影像在分析後並不會保存下來,所留下的只會是統計完成的數據,確保私隱。

其他較常見的電腦視覺應用例子還有:


道路車流量觀察

運輸署可以利用安設在道路旁的攝像頭,實時監測實時道路狀況,例如車流量觀察、車禍或特別事故發現、交通擠塞狀況等。這些鏡頭的記錄重點在於車輛輪廓和道路的整體使用量,以用作計算車流量和路面狀況。有了這些數據,運輸署便可以及時透過手機應用程式、網頁及電台發放最新的交通資訊予各道路使用者,方便他們選擇最適合自己的行車路線和出行方式。

火災偵測

在室內,我們可以利用天花板上的獨立火警偵測器(又稱煙霧探測器)偵測火警的發生。但當煙霧探測器響起警報,往往代表著火災已發生。煙霧探測器除了無法應用於戶外環境,亦無預防火警的作用。我們又該如何探測位於室外的火災或即將釀成火警的火苗呢?電腦視覺技術便可在這裡派上用場。運用閉路電視鏡頭和紅外線等裝置,這些鏡頭的記錄重點在於火災黑點和整體環境溫度,記錄實時畫面和物品溫度,再利用電腦視覺實時進行圖象分析,偵測出懷疑煙霧、火苗或異常高溫的物品。提示管理者及早派人巡查和及時通知消防單位,達至防火目的,降低大火釀成的機會,保障人命和財產安全。

在本文的尾聲,不如我們一起回顧一下開頭提及的「野豬事件」,若我們能應用電腦視覺及時發現到野豬的蹤跡,並將其引導回郊區,相信此類案例發生的機會會大大降低。通過本文中所提出的種種案例可以見,電腦視覺技術除了有助提升城市運作效能外,當中的實時觀察功能亦對保障城市及市民安全別具重要性。

在社交媒體上追蹤我們!

Facebook | LinkedIn | Twitter