“Measuring characteristics, like the distance between the eyes, the dimensions of the nose, the angle of the jaw. Ferret’s struggling to interpret data to create a template. Based solely on the multiple images, created by different people, the chances are slim that we’ll get an accurate portrait. But since you ask, we’re looking for face-to-face analogies. Are the eyes similar enough? The bridge of the nose? The lips? Teeth? The things that are hardest to cosmetically change. The program is processing distinct facial regions, encoding parameters from the rectified images, and attempting to establish a norm.”
“And it works?”
“Sometimes.”
“Oh, great.”
“Well, that’s the truth. I’d say Ferret works pretty well provided the materials at hand are good.”
“Like a photograph?”
“Even then, it’s no slam dunk,” Touch allowed. “It’s best when the subject’s face occupies the whole image. Unless they’re holding a number in front of their chest, you tell me how many bad guys will pose for the camera?”
“Back to my point. I got it.”
“No, you don’t. Not yet,” the FBI expert complained. “Ferret recognizes the improbability of that. It’s programmed to examine more complex images and sort out the extraneous, re-focusing on potential faces within any given field. Using a neural network-based face location system, it locates possible faces within an un-composed image.”
“Then it works,” Roarke said, hopefully.
“Depends on your definition of works,” Parsons snorted through a laugh. “One of the first major applications was Super Bowl XXXV back in 2001. Everyone entering got their punnum’s scanned. Mind you, no one realized it was happening. But the FRT cameras clicked away as they went through the turnstiles. The digital pictures were fed to computers, which looked for possible matches with known criminals. Note that I said criminals. Boy, we live in a different world now,” Parsons added as an aside. “Anyway, the software flagged nineteen individuals. Some were just petty thieves. It picked out a ticket scalper or two. Most of the rest were just false positives.”
“False positives?” Roarke needed some help with the explanation.
“Yeah. Falsely matching innocent people with database photos of perps. And then there’s the problem of the reverse-false negatives.”
“Not catching people even when the picture is in the system?”
“Bingo. That answer your question about it working?”
“Sort of. But it is an aid.”
Parsons nodded agreement. “For me, yes. Some others out there wouldn’t necessarily agree. There’s a big debate on its use at airports. Concerns about the competency of Ferret delineating darker-skinned people in bright backgrounds. The effect of adverse lighting. Problems when there’s high red content behind the subject. Whether glasses throw off the analysis. Scars. Tattoos. Head-on shots versus profiles. Everything matters.”
“But now? Is it reliable now?” Roarke counted on an affirmative answer.
“Look, Roarke, with just sketches, you’re not going to get what you came in for. There’s no way to create a reliable extrapolation without at least one authentic picture. You realize the CIA and the Bureau don’t have good pictures of most al-Qaeda. Even when we access an archival passport photo from Interpol or other agencies, the programs still need further development to handle the aging process. You and I have been through that already.”
Roarke nodded agreement.
“And the technology is still fooled by weight gains and beards. That’s why it’s taking so long. Too many geometric variations. Too much differentiation in age. Too much…”
Starting at the top of the screen, an image began to render.
“You were saying?”
“Well, the program reduces measurements of human faces to mathematical formulas or patterns in the database. New software out of Stony Brook detects minute patterns of muscle movement in a smile. It’s becoming one of the best indicators. Imagine that. A smile can be like a fingerprint. We call it a ‘smile map.’ And there are other facial landmarks that can come into play,” Parsons continued more humbly. “Apparently…” he paused. A definite soft composite picture was resolving on screen.
Roarke smiled. “Yes?”
“Apparently, it detected enough landmarks to achieve a robust divination.”
“In English.”
“A crude prediction.”
“Crude? Crude sounds real good to me right now,” Roarke said.
“It’s the best you’ll get until…”
“I know. Until I hand over a real photograph. But this is going to help.”
“Will you leave if I sharpen this up?”
“On my honor.” Roarke held up three fingers: the sign of a scout promise.
“Why do I think you’ll be back?”
Roarke stood and slapped Parsons back. “Because you know me.”
Parsons waved him away and typed a prompt. The computer acted as if it had been given the equivalent of a stirrup to the hindquarter. Seconds later, a new image began to render. An almost photographic face gradually took shape, growing more real with definition. Colors and shading began to give it character. The chin was as Roarke remembered. The eyes deeper and thin. The rest was familiar, yet different.
The computer finished its work, and Parsons immediately saved it to the hard drive and printed a copy.
Speechless, Roarke studied the work. Depp looked to have an almost military quality. The composite depicted a man in his early 30s, Caucasian, computed with a muscular, chiseled face, a thin lower lip, an undistinguished nose, high cheekbones, thin eyebrows, close-set ears, short brown hair, and the cold eyes Roarke remembered. All in all, the likeness appeared very similar to Roarke in facial sculpture. Except for the scar under Roarke’s chin and his short brown hair, they could be brothers.
“Well?” Parsons asked, fishing for a compliment.
“I don’t know. I really don’t know. It’s different.”
“Of course it is. As you said, your Mr. Depp is a master of disguises. But this may be as good a look behind the mask as you’re going to get until you’re face-to-face again.”
Roarke peered into the screen. “Some of it seems right. Some of it….”
Parson’s interrupted the thought. “Now let’s see if he made the mistake of standing in front of a camera somewhere.” Parsons saved the image to another program and typed in a new command. This time, Roarke let the photo expert continue without comment. Ninety-one seconds later, Ferret delivered the answer.
“I just accessed all of the known terrorists in the memory, along with state-by-state motor vehicle license photos, FBI records, newspaper archives, military IDs, and dozens of other sub directories. Here come the results.”
The computer image reduced to one half the page, with the picture on the left and data on the right. Parsons read the analysis.
“Your man is a possible match to, let’s see….” He highlighted a single line of text. “7,451,209 other subjects worldwide.”
“Oh, fuck!” Roarke swore.
“And that’s assuming he’s even in the damned database. Wanna bet he isn’t?”
Chapter 5
He was waiting for her. Lynn Meyerson had already circled the Cheviot Hills Park and Rancho Golf Course once, a run of about three miles. Olsen planned to strike on mile six of her second lap.
He watched her still-powerful, long strides come into view again. She had circled the golf course, cut back into the park, and now darted across the grass near the parking lot. Her run took her between two baseball fields where Little League teams played. Olsen stood up from his park bench and stretched. Maybe she would have sprinting power left, he thought. He better be ready himself.