June’s Geointelligence Insider article on Jack Maple was the human interest article. One of the readers of June’s article had the opportunity to meet Jack Maple. I appreciate the feedback. This month’s article is based upon the recommendation of Geospatial Solutions Managing Editor Tracy Cozzens to cover the technical side of GIS and crime fighting.
Recap
Fighting crime with GIS sounds simple enough — map where the crimes are happening and where the bad guys are and send in the cops. That would be a gross over simplification. As always, there’s more to the case.
The first CompStat was founded in the pre-internet days of 1994 on a Commodore 64, harkening back to the days of 128-MB floppy disks and MS-DOS, a Jurassic period of computer evolution that marked some of the first steps into crime fighting’s digital era. The graph above is the current CompStat 2.0 from the NYPD. It is a GIS-based system, interactive, user-friendly and available to the public. Take note of the highlighted number.
Since CompStat was introduced, crime has fallen precipitously. As of this writing (Aug. 25, 2018) New York City has 183 reported murders year to date. By comparison, in 1990, prior to CompStat, there were 217 murders per month on average. More murders were committed per month in 1990 in New York City than what it will experience in all of 2018. Murders have dropped nearly 90 percent. In other words, nine out of 10 who would otherwise have been killed are alive to return to their families, parents, classmates and colleagues, and friends. The difference is staggering, providing tangible proof geospatial science is a benefit to humanity.
The arsenal of geospatial applications available for the crime fighter is enough to make any superhero envious. The list of these high-tech, integrated intelligence systems push the limits of science fiction.
The underlying strength of these systems is the robust GIS/GPS platform they are built on. Security cameras are geospatially connected to the network with dynamic mapping capabilities. This allows the surveillance video to be overlaid on GIS software in order to interact with more information and create actionable intelligence.
Cameras use a host of software algorithms that are able to recognize aggressive behavior, patterns, anomalies, change detection, biometric features, objects and text. Systems can integrate real-time information like social media feeds as well as live video, including facial recognition software scans for wanted individuals in real time. This information is shared with police officers in the field.
Police cars are becoming mobile command centers outfitted with a suite of sensors, and will eventually include drones. Police officers wear smart glasses augmenting information about who they are looking at in an intelligence and location-based context.
The police officer’s belt is Bluetooth enabled, connecting all these devices, as well as monitoring the officer’s vital signs. The officer’s gun is also Bluetooth enabled, reporting when it is drawn, the direction it is pointed and if it fired. The gun also comes with a chip for tracking purposes and to ensure only the officer it belongs to can fire the weapon.
The vest-worn camera is woven into the seamless geospatial network of sensors and records the officer’s experience from a first-person perspective.
Imagine this scenario. A crowd gathers at an intersection triggering an anomaly detection sensor due to the number of people gathering in that location at that time. Out of the thousands of security cameras being monitored at the Command and Control Center (C3), this video scrolls around and flashes yellow, bringing attention to the possible situation. Other mounted cameras that have that intersection in the field of view automatically align along the edge of the main security video projecting their imagery onto a 3D data model of the area. Police officers in the field nearest to that location are simultaneously alerted. No action is taken at this point except the police begin heading in that direction. Facial recognition software scans the video images for faces of known suspects. Social media and texts scroll next to the video and geospatially link to those in each frame of the video. Colored sentiment indicators showed levels of concern. Boxes outlining people in various colors correspond to threat levels determined by datamining multiple databases. Semi-persistent motion trails lag behind each box showing the speed and direction of people in the video. Pattern identification looks for convergence, divergence and synchronous movements.
Video analytics identify several people converging on a car that just pulled up. The license plate reader linked to the security camera reports the car as stolen with two traffic violations. Based on this preliminary information, the situation is elevated. A police officer is dispatched but before arriving, a drone launches from the police car outfitted with a true color camera and a stereographic infrared camera. The stereographic imagery pair is streamed live to the police officers entering the area of interest through their smart glasses and to a team of imagery specialists at the C3. The video analytics of the police drone are seamlessly integrated with the security camera videos focusing on the car and the individuals as it arrives on scene and surveils the area. Object recognition identifies three possible weapons on the persons of interest. The boxes around those individuals turn red. They are tagged for persistent surveillance by all security cameras in the area. The order is given to apprehend them for probable cause. More police officers are called in and before they arrive they know who they are looking for, where the person is located, and that they may be armed and dangerous. In less than a minute, the police arrive. The suspects flee. The drone follows one of them up the street into an alley. Two of the officers pursue him. The other two suspects jump into the car and drive away. License plate readers and security cameras track the car on a map showing the vehicle’s route and speed with corresponding real-time video as the vehicle passes into view of each camera. As the vehicle travels south a police officer steps out from a cross street and shoots an electromagnetic dart into the speeding vehicle, disabling it. The police officers approaching the car shine a disorienting laser light weapon called a dazzler at the suspects, preventing their eyes from focusing. The occupants are apprehended without incident. They are searched for probable cause and arrested for carrying handguns without a permit.
The other suspect fled on foot. The drone followed him relaying live imagery to the police officers’ smart glasses. Their smart glasses showed a real-time map of their locations and the suspect’s. They cornered him in a fenced area. Guns drawn, the smoky red light of the laser cutting through the air pointed at the suspect. He surrendered. No gun was found on the suspect but the drone video the gun being thrown into a dumpster. One of the officers went back and retrieved the gun.
Gun traces were run on the three confiscated weapons and one was identified as stolen, matching a description of a gun used in a recent homicide. One of the suspect’s fingerprints match those found on shell casings at a nearby location reported by gunshot acoustic sensors. Based on this information, there is probable cause and a tap and trace is approved electronically by a special task force judge. The phone records of the three suspects are searched linking them to the el Diablo gang. Several unknown numbers are also in the call logs. Those numbers are added to the case file to be investigated later.
Only one of the suspects has a known address. The other two have no known location. Activity extracted from phone records show their whereabouts over the preceding days pin pointing their main locus of operation. Search warrants are issued. Within hours of arresting the suspects, the locations are raided and searched. Officers discover a cache of weapons, drugs, laptops and other useful information.
Everything described above is already available — it is only a matter of time and money. And, if Dubai is any indication of things to come, police could soon be arriving on hoverbikes.
The police arriving within minutes is key to the success of preventive policing. Time saves lives. The goal is to intervene before crime happens. But how is it possible? Before answering, let’s look at some numbers.
By the numbers
In 2017 almost 84 percent of the population of the United States was considered urban residing within 106,400 square miles. The Bureau of Justice Statistics reports there are only 758,854 sworn officers in the United States. Maintaining the same 84 percent ratio as the population means only 634,847 officers cover those urban areas. Specifically, it breaks down to six police officers per square mile. It is one police officer for every 431 residents except that police, like all of us, work 40 hours a week, have days off, take vacations, etc., so, only one out of every six police officers is on duty at any given time. That is one police officer for every 2,153 residents; however, police often operate in pairs, so 4,306 residents depend upon two brave souls to protect them from danger.
Victims of violent crime are 2.1 percent of the population. In a sampling of 4,306 residents that equals 90 victims of violent crime every year. In the top 10 cities it is far worse. Police officers have an incredible responsibility placed on them and they rightly deserve our praise, support and respect for the dangers they face every day.
Why not more officers you ask? Police protection comes at a cost of $100 billion annually. Our relative safety is not cheap. Crime is a huge expense. Jails, trials, public defenders, prosecutors, judges and incarceration all cost money. Safety is expensive. Budgets are stretched thin. The answer for increased safety and security isn’t more police. The answer is integrated and intelligent technology systems leading to increased efficiency. Technology has benefitted most other professions. Now, the field of law enforcement and crime prevention are benefitting. Cost is the driving need. These efficiencies are being realized on a grand scale. Making matters more urgent is the worldwide mass migration as populations move towards cities. It is imperative to manage crime now rather than later.
Enter predictive policing — putting the power of open data, cloud computing, machine learning, geoscience and artificial intelligence in support of law enforcement and prevention. Basically, cities are broken up into grid patterns, typically 500×500 feet. Within each grid, crime data is compiled using multiple factors and resources, such as historical data, 9-1-1 calls, recent crime reports, and residences of known offenders and parolees. Even considerations such as the time, day of week, celebrations and cyclical events are taken into account. Information derived from security cameras, license plate readers, social media and financial transactions help the algorithm. The algorithms take into account information collected by authorized wire taps, call logs and other confidential sources. The goal of the algorithms are to include all available resources to develop the most complete and reliable dataset upon which the heatmaps base their probabilities. This helps police departments allocate their resources, know what to prepare for and, most importantly, know where to be to protect the public at large.
Police tighten their patrols around the hotspots. Throughout their shifts those hotspots are subject to change depending upon new data. Mobile units simply focus their patrol efforts accordingly. Once a threat is reported, automated navigation routing systems show police the fastest route to the incident and their expected time of arrival. Officers continue to receive intelligence about the incident while en route to anticipate the situation prior to arriving. Knowing where the areas of highest probability are expected to occur focuses non-human assets too, such as geofencing the areas of interest and monitoring more closely for key indicators. This technology is not too different than numerical weather forecasting models predicting what and where weather events will occur in the next hour, three hours, six hours and so on. Numerical models continue to evolve making forecasting more and more reliable. And, although the past does not predict the future, it is a strong indicator. The disclaimer would be similar to the ones most have seen before, “Past performance does not guarantee future returns.” Sometimes preventing a crime is saving a life, sometimes it’s protecting property and almost always it is stopping someone from doing something they will later regret. All crime cannot be prevented but for every crime that is prevented there is a family spared from tragedy.
Preventive policing does more than help keep communities safer. It improves economic viability. Crime has an inverse relationship with a community’s vibrancy. As crime increases, prosperity decreases. Real estate values go down, the tax revenue goes down, employment opportunities go down, and safety, happiness and well-being go down. Crime is a societal disease. Reducing crime reverses those affects. Home values, employment, affluence and the quality of life all go up, which correlates to increased tax revenues. Thus, reduce crime and the city’s revenues increase. That means politicians can divert money into other programs to benefit the citizens. For these reasons there is bi-partisan support for computer based policing.
If you do the research you will see opposition efforts against artificially intelligent systems to fight crime, but those opponents are not well supported. Communities want to feel safer. Politicians want to be able to say they are using the latest technologies to keep the community safe. Companies want to prove their systems work in decreasing crime and capturing criminals. Crime prediction causes the greatest concern because it borders on Minority Report, but it is the echo of Jack Maple and William Bratton putting police where they need to be to support the people they need to protect. It is the essence of community based policing.
This article only touches on the front side of GIS and law enforcement, but there is another world on the back side piecing crime scenes together with forensics in artificially replicated environments. That too is a fascinating topic to explore.
Do yourself and your neighborhood a favor. Thank the police officers in your community for the job they do. They are foundational to the fabric of our society.
