State of Surveillance (Cameras and Software)

By Mareike Ohlberg and Jessica Batke

This article is part of a larger investigation of the PRC’s surveillance apparatus, including a feature article, a number of sidebar articles, and supplemental materials, all originally appearing on ChinaFile. The text below does not include illustrations and charts that are part of the full article.

To reach their goal of monitoring all of China’s key public spaces, local authorities need to buy a lot of cameras and other technology. The approximately 76,000 surveillance-related procurement notices ChinaFile analyzed offer a glimpse into the kinds of hardware and software used throughout China to monitor public spaces.

As they have built out surveillance systems in their jurisdictions, local authorities have purchased many different types of cameras to address a range of goals, procurement notices show. In general, the most common types of cameras, used to realize “global coverage” of all key public areas, are high-definition (HD) bullet- or dome-shaped networked cameras with resolutions between 2 and 4 (occasionally up to 8) megapixels.

The definition of “global coverage” differs by location. A notice from the town of Shishan in Guangdong province sought cameras to help it achieve “30 percent coverage” of key public spaces, for which it would need 30 cameras for each of the town’s 330.6 square kilometers. This means some 9,900 cameras. By the officials’ definition, to achieve full coverage, Shishan would need more than 33,000 cameras, or 100 cameras per square kilometer.

How does this compare to, say, the city of Chicago, known for its surveillance camera build-up in recent years? As of 2018, Chicago had 30,000 government-run cameras, spread out over an area of 600 square kilometers, or 50 cameras per square kilometer. Elsewhere in the American midwest, the city of St. Louis runs about 400 cameras in the city proper, or about 2.5 cameras per square kilometer. Notably, St. Louis also connects about 200 private camera feeds to its Metropolitan Police Department headquarters, which brings camera density to 3.75 cameras per square kilometer. The distinction between public and private cameras is a salient one when comparing China with other locales, as many estimates of the “most surveilled” places in the world include private cameras, which are not necessarily linked up with police networks.

Chinese government officials sometimes purchase specialized cameras, whose purported capabilities offer clues as to local authorities’ aims for specific areas. These are often “smart” devices: devices integrated into systems that can algorithmically ferret out purportedly suspicious behavior, link actions in the real world to a trove of personal data, or even attempt to “predict” who might be dangerous. For example, officials want some cameras, installed in places such as government offices, hospitals, schools, and financial institutions, to measure “population density.” These monitor “the flow of people . . . in real time, and [automatically] trigger an alarm in case of major events such as abnormal crowd gathering.” Authorities in Zhejiang and Guangxi provinces have also sought cameras that offer alerts when they detect certain behaviors, such as a person or vehicle entering or leaving a certain space, a person pacing back and forth, abnormal noises, border crossings, or any unexpected changes to a background scene. (It is worth noting here that, as long as their resolution is good enough, authorities can also use images from more standard cameras as fodder for such automated analysis.)

To manage areas with large, disorderly flows of people, police seek “integrated intelligent tracking” cameras that can both monitor fixed targets and track moving targets. Some procurement notices also seek the ability to reconstruct a person’s path. A user could select multiple images of the same person from a “snapshot library,” and the system would automatically draw the target person’s route on the system map according to the time and location that the facial images were captured. In addition to “smart” devices, public security authorities employ “panoramic cameras” so that they can record a wide-angle view of an entire scene while still being able to see details clearly. Surveillance systems can also include high-altitude cameras, sometimes with thermal imaging capabilities, that overlook an entire area. In Yuepuhu (Yopurgha) county, Xinjiang, for example, local officials wanted to set up nine high-altitude “thermal imaging monitoring points,” one in each of its seven townships and two in the county seat, presumably to detect body heat.

Traffic or highway cameras serve multiple purposes. They are, of course, supposed to pick up on actual traffic violations, such as whether or not a driver is on the phone or is wearing a seat belt. But authorities also use them to closely monitor who enters and exits their jurisdictions and to track down specific individuals. Some surveillance systems, like the one sought in Foshan’s Shishan township, are trained to recognize suspiciously similar license plate numbers, because, as a procurement notice from Shishan put it, “many criminals frequently use stickers to change one or several of the digits or letters on their license plates in order to evade monitoring.” Many authorities install cameras at specific distances along highways to identify license plates and car colors, among other aims. One notice from Liaoning province specified that their equipment needed to be able to identify more than a hundred car brands and several thousand sub-brands with 97 percent certainty during the day and 93 percent certainty at night. The same notice mentioned a camera that could simultaneously track pedestrians, non-motor vehicles, and motor-vehicles and “connect human faces with bodies, license plates, and vehicles.”

A range of Chinese companies manufacture the equipment detailed in these procurement notices. These include Hikvision, Dahua Technology, SenseTime, Megvii, Uniview, Yueshi, Sunell, Digital China, and Ausee. Equipment can also come from Chinese subsidiaries of foreign companies, such as the Shenzhen-based subsidiary of New Jersey-headquartered Infinova.

Procurement contracts usually are not awarded to foreign manufacturers directly but to third parties who purchase the equipment and provide it to the government. Manufacturers (both foreign and domestic) are unlikely to install this equipment themselves. This means that technology company names, including foreign suppliers of technology, do not generally appear in initial procurement notice announcements, although they are sometimes buried in the lengthy addenda that some notices contain. For researchers trying to understand how domestic and foreign companies’ products feed into China’s surveillance regime, this complicates the gathering of such information.

Local governments install different surveillance technologies as each successive wave of surveillance campaigns reaches their jurisdictions. Therefore, the technological capabilities (or lack thereof) sought by government officials embody a moment in time: How much money did a given office have on hand to spend on these devices? What capabilities did it deem most vital? And what capabilities were even technologically feasible at the time? As the next national surveillance campaign rolls through the country—whenever that may be—the equipment will no doubt be far more advanced, but the choices and trade-offs they represent will likely remain as geographically varied as they are today.

Jessica Batke and Mareike Ohlberg