I started my second Silicon valley startup (using the money from selling my first one).
My business idea (at that time unique) was to provide set-top and mobile video entertainment systems... basically a disk drive (or two) with a built in player hardware / software in a compact platform that one could move from the house to the car (minivan). My idea was to have every Spongebob or Pixar film or Disney film ever made (plus video games, etc) in a single system that had an easy carrying handle and docking station for both the home and the car. One simply pulled the box out of one cradle and carried to the car and all of the kids (mostly thinking of kids) movies could go on that camping trip or whatever.
Anyway, that got me into the world of mobile digital video.
And then the cop car camera people came knocking on my door. They wanted to move from tape to digital recording for cop dash cameras (and, more important) the back seat cameras. The reason was one of liability... municipalities did not want to defend cops that were actually abusive to prisoners or to payout to accusations of police brutality when none actually occurred.
They wanted the following:
1. A digital system that would start recording automatically upon any number of trigger events
a. turn on lights or siren
b. remove shotgun from dash holder
c. remote microphone activated
d. back door open.
2. They wanted it to transfer the shift video wirelessly from car to HQ
3. They wanted time, date, and GPS information recorded.
4. They wanted the video to be acceptable in court (compression had to be one accepted by courts)
5. They wanted the entire "event" recorded, and not editable.
This wasn't necessarily the police departments, but the city governments that wanted all of this.
I started thinking about the problem (#2 was technically difficult at the time, especially when coupled with #4)
My first idea was the the camera(s) should be "recording" all the time... but we only save the recording from like a minute before a "trigger event" from #1 to a few minutes after the event is complete (police indicates event over). That way the court might see what caused the police to do the trigger event in the first place.
My second was to couple this video system with a OCR software and take a picture of every license plate the camera "sees" (OCR the record, take a still picture in jpg, tag it with GPS, time, date, etc). Then, all of the OCR info and GPS info is stored in a huge nationwide database... and should there be a warrant at some point in the future, the database can spit out all of the locations and pictures of the license plate over years of observations. That would allow the police to pinpoint likely locations, number of occupants, changes in appearance of the car and driver(s) over time, etc. I was thinking things like "amber alerts" and how this could help police track down a kidnap victim quickly.
The amount of data is both "staggering" and trivial... it turns out that the observation records are only about 50KB per observation... x 8000 per shift (my estimate), x 100 police cars per shift for a medium sized city... 40 GB/day for a medium sized city or 100 TB for 7 years worth of observations... 4 TB disk drives are something like $300 a piece now... a modest 100TB storage system is only a few thousand dollars.
I didn't patent the idea nor did I share it with the cop camera companies that I was dealing with.
I just couldn't bring myself to making probably millions of dollars ushering in the era of Big Brother.
I knew it would happen eventually (and has now).... but I just wasn't willing to do it then (and under Bush, there was plenty of money floating around Silicon Valley from the newly formed DHS for just such "fight terrorism" projects).
= new reply since forum marked as read
