Quality-Checking the Camera-Detector Databases

How accurate are Web downloads of red light and speed camera locations?

by Craig Peterson

Last updated 7/1/2012

Their location databases allow GPS-enabled radar detectors to warn of speed cameras. But only if someone remembered to add the data.

Also see our test of Internet/smartphone community-
based apps Escort Live and Cobra iRadar.

The colorful acronym GIGO is credited to George Fuechsel, a fifties-era computer programmer. Way back when ENIAC was cutting-edge technology, Fuechsel presciently cautioned, "Garbage in, garbage out", i.e., the quality of the information produced by a computer is wholly dictated by the quality of the information entered into it.

We wondered if Fuechsel's cautionary aphorism also applies to the heart of every GPS-enabled radar detector: its digital list of camera locations. So we set out to answer two big questions: Exactly what constitutes a superior location database? And, which one is the best?

One crucial task we quickly identified: keep it up to date. New cameras can literally appear overnight and adding a site weeks or months after it becomes active simply isn't good enough. Days matter. For example, during a GPS-enabled radar detector review last year, on a Friday afternoon we left Phoenix en route to San Diego. Our detector reliably alerted to each of the seven speed cameras that hovered over I-10 traffic like peregrine falcons scouting for prey. Four nights later we returned—only to be greeted by the blinding flash of a camera. There was no warning; the new site had become active since our departure.

Each company swears that its database receives constant updates, faithfully adding new sites and culling outdated ones. But that's a herculean task, entailing endless hours of research and a site visit to verify each camera. Keeping the database current and comprehensive is an elusive, constantly moving target. Don't expect perfection; it isn't going to happen. (If NORAD, with its near-limitless resources, can't guarantee that it tracks every aircraft in the sky, how can a handful of guys in rental cars be expected to keep tabs on every photo enforcement camera in the nation?)

We quickly learned why there's been no independent test of these databases: it's a formidable challenge. And the vendors certainly aren't eager to help. Each is so tight-lipped about their proprietary data that breaching the security of Bill Gates' office PC would be child's play in comparison. Our only solution was to employ the same data gathering techniques used by the vendors—and compile a database of our own.

We first narrowed the project's scope to four states—Arizona, California, Colorado and New Mexico—and zeroed in on six large metropolitan areas within those. Next, we compiled data from an array of public sources—news reports, municipal Web sites, law enforcement blogs, even the camera vendors' own press releases. Although camera enforcement contracts should be of public record, some towns purposely conceal their agreements with the vendors, forcing us to issue formal requests under state Open Records laws.

The process took months but finally, we had our database completed. Then came the labor-intensive part: we placed a test fixture in our target car, installed the radar detectors and began driving to 128 camera locations we'd identified. Arriving in an area, we would check into a hotel, update each detector online and then begin visiting each target location, checking sites for operational cameras. (All radar detectors were updated daily.)

We verified 96 working cameras in all. Another 11 sites had no cameras, poles or wiring, all had either previously been photo-enforced and since abandoned, or had been announced as future sites, but no cameras had yet been installed. Two sites were under construction.

All of the remaining cameras were phantoms, many of the locations having been supplied by Web sites like photoenforced.com (Phoenix map shown below; each icon represents a camera) and others that offer this information either for free or at a very nominal cost.

But we soon learned there's a hidden price to be paid for cheap data. Most sites like these rely on enthusiasts to report new cameras and it became obvious that many helpful contributors can't distinguish between a traffic surveillance video camera and a red light camera. Nor are these reports vetted, judging from the substantial percentage of claimed sightings that proved to be erroneous. This again illustrates a phenomenon intrinsic to the Internet: it's awash with free information, much of it wrong.

When we found a camera open for business, with one detector powered-up at a time we approached the location from one mile away, waiting for an alert. The particulars of each were noted, including whether its GPS coordinates accurately depicted the camera's location. (Incorrect coordinates mean the detector might alert either far in advance or, worse, long after the camera has been passed.)

Test Results

We found the Aura database precise in its location-marking, allowing a Cobra GPS-enabled radar detector to reliably alert at the appropriate moment. It was also very reluctant to issue phantom alerts to cameras that are either nonexistent or which are merely traffic surveillance units, an affliction that plagues many databases. This would indicate that Aura reps are dispatched to verify locations, a commendable practice. We noted, however, that the Aura system trails the Escort Defender at deleting outdated locations, making it more prone to unwarranted alerts of cameras often removed years previously.

The Cobra Aura database showed excellent accuracy against cameras that had been active for a few years. But newer installations seemed to be problematic: this database appears to lag noticeably in receiving timely updates. Analyzing our data, we confirmed that without exception, the cameras it missed had been operating for less than 18 months. For example, each of Tucson's speed cameras had popped up within the past 120 days: the Aura database missed all 10 of them. At the time of our visit, metro Denver's cameras had all been in service less than 18 months and the Aura missed 13 of the 13. Omissions like these greatly affected its average rate of accuracy. When we crunched the numbers, we found that it had alerted us to only 66 percent of the operational cameras.

A Cobra spokesman expressed surprise at this news and suggested that either the detectors were somehow faulty or their downloaded GPS updates may have been incomplete. So although it seemed unlikely that both units had failed, to be fair, we acquired a second XRS 9960G and revisited a representative selection of the sites. This one performed identically, suggesting that the issue is data-related, not a hardware problem. We couldn't help but notice that allthough the Cobra Aura Web site claims to be updated every 12 hours, on two occasions this year we logged-on units after a months-long absence only to learn that our database versions were the most current available.

In contrast, the BEL and Escort Defender database alerted us to fully 95 percent of the cameras. This is weapons-grade camera protection and about as good as this type of data-delivery gets, at least without enlisting some psychic assistance. Defender also mapped camera locations with unerring precision, enabling the BEL GX65 and Escort Passport 9500ix to reliably inform us about the particulars of each camera encountered and accurately depict its proximity.

The Defender database proved conscientious in deleting obsolete camera locations and was very quick to add new ones. Weekly updates are provided and each helpfully denotes which states are affected, saving users from unnecessary downloads. The BEL (Beltronics) GX65 and Escort Passport 9500ix ship with databases already loaded, ready for use. A one-year Escort subscription to online updates is $19.95; a three-year package costs $29.95.

Camera-detection GPS Alternatives

Although they're not included with our GPS-enabled radar detector review, we also evaluated some other warning devices intended to counter the photo enforcement menace. These include red light camera detectors, a Garmin Nuvi 350 satnav and a mobile phone app called Trapster.

Another product launched too late for inclusion is Escort Live, a promising meld of the Internet, GPS and a smartphone app. Combined, the system lets users broadcast warnings to others and automatically sends out alerts to various threats.

Trapster is a free service that uses the amazing Google Earth to display enforcement threats in several countries. The app works on half a dozen mobile phone models as well as Garmin and TomTom portable satnavs. It supplies warnings of all types of photo enforcement including red light and speed cameras, plus photo radar vans. Unlike the Defender and Aura systems, Trapster also furnishes Live Police Traps alerts, user-reported sightings of speed enforcement activity in progress. These generally are considered to be officers spotted using lidar or radar, or those thought to be on the lookout for traffic infractions. Another alert, Police Often Hide Here, reflects locations where users report that traffic enforcement frequently occurs.

Trapster is a democratic forum and users are invited to vote on the accuracy of reported threats, in turn generating a "confidence rating" of a location's veracity. Clicking on a location provides a data box with details, even a street-level view of the location, sometimes allowing a peek at the red light or speed camera. Or not. We found some recently reported camera locations to be devoid of any hardware more threatening than a stop light. Many of the areas tagged as frequenty-enforced hotspots could well have been the scenes of previous traffic stops. But the likelihood of stumbling across a traffic officer there subsequently is too low to be considered troublesome.

We're equally confident that a high percentage of the Live Trap Alerts were sparked by the sighting of a police patrol vehicle, most of which are not toting speed-enforcement hardware. These officers are certainly capable of writing traffic tickets, but their primary focus lies elsewhere.

Trapster quite accurately depicts fixed cameras, although a number of phantom camera locations are included as well. This also goes for photo radar vans, most of which move to new locations every three to four hours. This makes even near-real-time reports of questionable value since mobile enforcement simply moves too often to be predictable.

Other drawbacks to Trapster are its reliance on the cellphone and the need for frequent interaction. Users will find themselves deeply engrossed in manipulating the various pulldown menus, navigating about Google Earth and studying the information displayed. Even veteran texters and accomplished multi-taskers may find themselves steering with their knees while devotng a significant amount of attention to the endeavor. The risks are high enough that during our testing, eventually all the drivers refrained from using the service while in motion, allocating the task to a passenger. The heavy reliance on visual alerts, the iffy method of user interface and the prevalence of unverified data makes Trapster-style databases perhaps best suited to low-risk geeks with an excess of time on their hands. These same limitations also apply to the Cobra iRadar, a combination of a low-end radar detector with an iPhone or Android phone that we found to be of marginal value.

We also briefly sampled Trapster on our Garmin Nuvi 350. This nav system permits a more stable mounting location but otherwise exacts the same limitations: lots of user interaction, made even more tedious by its comparative lack of controls. It works acceptably well, but we'd advise caution when using it.

There's one other limitation to these gadgets: none detects radar or lidar. City fathers quick to embrace photo enforcement often give the nod to radar vans as well. Our own recent study proved that like speed cameras, these have absolutely no effect on speeds, but they certainly generate revenue. They're spreading faster than the swine flu and unless you're packing one of the select few radar detectors proven capable of countering them, you're definitely at risk.