It’s becoming a fact: The next police officer that you come in contact with will most likely be wearing a body camera. Considering that there are at least a million officers patrolling our streets, you can only image the impact that this massive amount of new content will have on business systems, operations, infrastructures, and personnel in just the next few months. The question is obvious – are we going to be able to control body camera video, or is it going to control us?

Trying to answer that question ourselves, Vu Digital recently took on a project that involved analyzing body worn police camera video that could be potentially used for evidence by city and county prosecutors. Video evidence is important and can be critical in helping judges and jurors make the right decision on a verdict. It’s very important to know what’s in a video and how to get to it quickly if it’s deemed important. We wanted to prove to our law enforcement clients that we can understand video by simplifying the process of identifying, tagging, editing, and storing large amounts of video. Kid stuff, right?

It’s only kid stuff if your kid is good at math because there’s a significant problem if you want to get results timely enough to actually be useful. Now we’re getting to the real problem. Any given police officer could record as much as eight or more hours a day of critical evidence (and we’re only counting when he’s recording something related to an arrest or investigation; there’s much more that could be recorded). Considering that officers usually travel in pairs, that’s fifteen to twenty hours per shift of video. We should also mention that patrol cars will also have a camera, two actually – front and back. That’s another fifteen to twenty hours, so now we’re up to thirty to forty hours of video for one crime scene in one day.

It’s becoming a fact: The next police officer that you come in contact with will most likely be wearing a body camera. Considering that there are at least a million officers patrolling our streets, you can only image the impact that this massive amount of new content will have on business systems, operations, infrastructures, and personnel in just the next few months. The question is obvious – are we going to be able to control body camera video, or is it going to control us?

Trying to answer that question ourselves, Vu Digital recently took on a project that involved analyzing body worn police camera video that could be potentially used for evidence by city and county prosecutors. Video evidence is important and can be critical in helping judges and jurors make the right decision on a verdict. It’s very important to know what’s in a video and how to get to it quickly if it’s deemed important. We wanted to prove to our law enforcement clients that we can understand video by simplifying the process of identifying, tagging, editing, and storing large amounts of video. Kid stuff, right?

It’s only kid stuff if your kid is good at math because there’s a significant problem if you want to get results timely enough to actually be useful. Now we’re getting to the real problem. Any given police officer could record as much as eight or more hours a day of critical evidence (and we’re only counting when he’s recording something related to an arrest or investigation; there’s much more that could be recorded). Considering that officers usually travel in pairs, that’s fifteen to twenty hours per shift of video. We should also mention that patrol cars will also have a camera, two actually – front and back. That’s another fifteen to twenty hours, so now we’re up to thirty to forty hours of video for one crime scene in one day.

Capturing video from a body cam is pretty simple: just turn the camera on and off when needed. Even storing the video is fairly straight-forward: most cameras can be linked to storage systems such as hard drives, or even a cloud services. Managing the storage can present a whole new problem, but we’ll get to that shortly. If you’re a police officer you simply rinse and repeat daily. If you’re a prosecutor or a defense attorney, you’ve got a problem.

Let’s assume that an attorney has access to the location of the police body camera video. In the case above there are 32 hours of video that must be reviewed for relevancy. There are 24 hours in one day, and only 8 of those hours are usually what any one person would be required and willing to work. If this is a case that requires a 24 hour turnaround of the video, there could be as many as four or five full-time people reviewing, tagging, and redacting this ONE day ONE crime video. You can multiply this by the number by a multitude of components to easily see the financial impact of introducing video evidence captured by body cameras into today’s judicial process.

Houston (and Philadelphia, and any other city), we have a problem.

Knowing that there HAS to be a better way, we had to take a different approach to solving the problem. We must find a way to reduce the review time by three, four, or five times thereby reducing the headcount required to complete these tasks without compromising the evidence.

Getting started on the project

Guided by the old Indian proverb “you can’t really understand another person’s experience until you walk a mile in their shoes” we decided to understand the pain of video review by experiencing it ourselves. The project that we were engaged in was with a prosecutor in a city of about 2000 police officers. They had three deployed body cameras on officers who were piloting the experiment.

Every day we were sent video evidence for arrests, homicides, and domestic violence incidents – about 100 hours during the experiment. All of this video needed to be reviewed for relevancy to it’s respective case.

Step 1: Automated meta-tagging

The video was ingested into our own product, called SMaRT (Storage, Metadata, and Redaction Technology), to create time stamped metadata over each video. This time stamped data included timecstamped speech (a full transcript was automatically generated with this process), time stamped face detection, and text extraction. This created a very nice index over every video which made it incredibly easy to find references within the video (such as gun, knife, kill, drugs, etc.)

This was automated so no headcount or labor was required.

Step 2: Reviewing

We designed and built an interface over the automated metadata so that a reviewer could see the video as a whole – speech, text, and faces in one interface. This design allows the reviewer to edit quickly and to skip parts of the video where there is no speech or nothing detected. We found in this step alone there was a reduction in manual effort of 30 – 40%. A lot of body cam video is still and there can be long intervals of inactivity. The result of not having to watch inactivity is a large amount of time saved in the review process.

Step 3: Composite reviewing

In the early example, there were 4 videos. Composite reviewing is the ability to review all four of the videos at one time based on a synchronized time clock. There will be a lot of redundancy of data, but there will also be many instances where one camera picked up audio or faces or text when another did not. The ability for one reviewer to review 4 videos and see 4 transcripts in one view changes the labor requirements completely.

Step 4: Redaction

The purpose of the review process in this project was not only surface video evidence for trial, but also to satisfy requests from the media or defendant. When video is supplied for public use redaction may have to occur to mute some of the speech or ‘blur’ faces and unauthorized uses of brands or personal information. Redaction, video clipping, and the ability to share information and store video had to be considered in this experiment. Many agencies will deploy solutions like this without considering the enormous impact that the mutation of one video into multiple clips will have.

Step 5: Present the evidence

When the 100 hours were completed, we were able to electronically send complete transcripts that included speech, face, text, and a few manual annotations that we thought were important. The video evidence was now in a form that could be utilized from a report, from a tablet, a laptop, a desktop, or even a smartphone. Prosecutors in this project are imaging the day when they are using their tablets and smartphones in the courtroom to help make their cases.

And the verdict is…..

We went in to this project knowing that video evidence will be critical in a trial and the timely production of the evidence in the form of data is essential. In the “time to review equals the length of video” formula for reviewing video, 100 hours would take at least 100 man hours (probably more). The cameras used here did not support time synchronization, so we were not able to use the composite review process, but we still completed the work in less than half of the time at a one person work week, or 40 total hours. The composite review would decrease that by at least half as much or more.

Policing body camera video is now as important and essential as getting a case to trial. Back to the original question – are you going to be able to control it, or is it going to control you?

Share This