As we turn the corner into 2015, I want to highlight a once futuristic technology that is already changing our lives in many positive ways: computer vision—training computers to investigate images and understand them the way the human brain does, but without human fatigue and at massive scale. In years to come, this technology will continue to help make our lives safer and easier.
People have long pointed cameras at a wide range of significant events to document and monitor them. Traffic cameras record images of routine moments like kids in crosswalks near schools, and cell phone cameras record seminal events like political demonstrations in oppressive nations. To date, we have captured more hours of video than humans have time to ever process, and often, significant seconds of footage sit on hard drives or get buried in mountains of visual memories never to be seen by human eyes.
That's where computer vision comes in. Computers can process and reconstruct video to make those significant few moments visible to us. They can also be trained to look for small anomalies or markers and tirelessly scan through massive databases looking for the seconds we want them to find. We’re just starting to see benefits from this innovative and constantly evolving image processing technology.
Many organizations—big and small—are deploying computer vision to accomplish commercial goals, like building self-driving cars, detecting shifts in the earth’s environment, and helping lifesaving services get to remote areas. Here are five ways computer vision is contributing to better years to come.
1. Computer Vision in Cars
It started with warning systems in cars in the early 2000s: a loud beep indicates we’re about to back into a garbage can, run up a curb or nick a parked car. Since then, more and more vehicles have rear-facing cameras to see objects before we get close enough to warrant the loud beep. Most recently, automakers are installing sensor systems that recognize images that are predictive of potential danger, like a car stopping in front of our car. Once the danger is recognized, the system automatically brakes to prevent accidents from happening. And in the near future, innovative self-driving cars will use computer vision to change the way we even think about transportation today. By these vehicles will detect and process the thousands of images a “vehicle” encounters en route to arrive at destinations safely and without accidents, likely without a human involved at all.
2. Seeing the Scene 360
A decade ago, news and reports came to us from one—or two—reliable sources that defined the “truth”. Information came from sensible locations reporters could get to or happened to be at when something significant took place. Now, with a cell phone tucked in pockets, almost anyone can be a reporter. These cell phones capture endless stills and video, creating millions of potential sources of news and “truth”. Among those who have benefited from this plethora of video reports are nonprofit agencies like Amnesty International.
Computer vision and image processing software help agencies like Amnesty gather valuable video evidence collected by global affiliates to track the humane treatment of people worldwide. These groups can better fulfill their mission by gathering real—and often real-time—video, which can be material that governments may not want us to see. Image processing takes video evidence and helps truth seeking analysts extract the precise, key information to combine and reveal a more comprehensive vision of what’s happening. The result: a new side to the story supported by irrefutable fact. We’ve seen this type of video reconstruction on shows like “CSI”; it’s a technology used by law enforcement everyday. Now nonprofit organizations use it to expose injustice or motivate justice seekers to pitch in and help with dire causes.
As we continue to see the evolution of true investigative, crowd-based journalism, organizations can use innovative technology to make better sense of video and images to create stories that account for as many perspectives as possible. Used in this capacity, organizations and governments have the potential to cultivate greater levels of justice, safety and cultural competency.
3. Flying Kites Making Maps Better
Lately, I’ve been impressed with a couple of grass roots organizations trying to update the world’s maps and help people in times of trouble. Their efforts are aimed at mapping unmapped areas, such as far-away refugee camps and remote areas hit by natural disasters. These efforts make directing humanitarian services to people in distress much easier, and the process helps provide more precise mapping information to countries that don’t have immediate and reliable access to it.
OpenStreetMap.org is a crowd-sourced map of the world created by volunteers doing exactly this: mapping unmapped areas. The data is free to use under an open license and offers a collaborative vision of ever-changing geographies. To get new aerial information about unmapped areas, non-profits such as PublicLab.org are using cameras suspended in kites to capture images and trace over them to define roads in OpenStreetMap.
Computer vision and geospatial image processing is helping nonprofits and their crowds of volunteers work more efficiently. Of course, mapping is an enormous and ever-changing undertaking, but with image processing, using open source visuals (from kites, drones, manned aircraft), we can process billions of images to create more up-to-date and accurate maps. New roads can be designed and traced, blocked roads can be corrected, and other geographic obstacles can be updated and disseminated quickly and with greater confidence and reliability.
4. Amber Alerts and Crowds of Eyeballs
When a child is missing, a crime is committed or a natural disaster strikes, time is of the essence. Search efforts for any crisis need to start immediately – the faster a person or moving vehicle can be located, the better the outcome. If an Amber Alert is issued about a late-model sedan that left a shopping center parking lot between 11:00 and 11:30 am and was subsequently spotted 8 blocks away heading south, computer vision's fast detection and tracking abilities can search through traffic camera and aerial traffic footage to identify potential vehicles (e.g., all cars that came from that parking lot and travelled to the location of the second spotting) for human eyes to review. The initial computer search significantly improves the efficiency for the expert analysts who are tasked with reviewing potential vehicles. This sort of “heavy lifting” that computers can accomplish to assist human effort can be applied to a variety of searches where time is of the essence and millions of images, or video time, can be eliminated in order to get to the answer. (http://youtu.be/HnwDpg2vvdg)
5. Predictive Analytics
What if we could use computer vision to “predict” events that affect public safety? If we know that the density and timing of traffic in one intersection has lead to pedestrian accidents, can we analyze video footage from similar intersections to predict which ones are more dangerous than others? And, can we innovate alternative solutions by changing the bus schedules or the number of seconds the “Cross” sign flashes? Likewise, if merchants in the area of the annual New Year’s celebration are upset about storefront access, parking, and traffic snarls during the festivities, could we review the footage from past years to identify the hot-spots and better address traffic flow?
We can start the exploration now: we have cameras in place. We have a whole new generation of aerial cameras (satellites and drones) that can provide data about almost any area. With traffic patterns, the bigger the dataset the better. The possibilities for making accurate “big data” predictions and finding solutions are endless. We can take video of a dangerous crosswalk, see what the true data is, and make changes that could potentially prevent loss of life.
We can make our crosswalks more secure, we can innovate safer cars, we can find a missing person faster, we can locate unmapped locations, and we can reformulate the value of video in every aspect of life. We are just beginning to understand the mighty effects of computer vision –to help make our lives better and more efficient. We can accomplish so many more extraordinary missions in our fast growing and fast moving world with this once futuristic technology. Now.