Technical Processing Record

Surveillance Footage Index

Casesample-001

Files processed2

Footage covered1.3 min

Period covered2026-08-11 09:15:00.000 – 2026-08-11 09:16:00.000 UTC

Events indexed6 motion

Report generated2026-08-18 06:36:01.952 UTC

Prepared byPolarity Management LLC

File integrity

Each file was hashed with SHA-256 before analysis. Nothing in this process modifies the source footage; the hashes below let any party confirm that the files examined here are byte-for-byte the files they were given.

ID	File	Size	Duration	Format	SHA-256
F01	front-door_2026-08-11_09-15-00.mp4	42 KB	40.0s	h264 640×360	c0dc60862b7226551b2f66d2d162e3d4de7f1de1ba074a12132ef8b7606647f3
F02	loading-bay_2026-08-11_09-15-20.mp4	42 KB	40.0s	h264 640×360	323a03f6010ad40cac50dd0cba9435efd47ea2abdfff812559d26e4e34b88121

How each timestamp was derived

Every wall-clock time in this report is the file's start time plus the offset into that file. The start time comes from one of three sources, and they are not equally trustworthy. This table states which was used for each file so that no timestamp is presented as more certain than it is.

ID	Start of recording	Source	Reliability
F01	2026-08-11 09:15:00.000 UTC	container metadata (creation_time)	high
F02	2026-08-11 09:15:20.000 UTC	container metadata (creation_time)	high

Detection settings, and how they were chosen

The motion threshold is not a fixed number. It is derived from each file's own distribution of per-frame change scores, because a camera watching a busy road and a camera watching a locked stockroom have different baselines. The figures below let the calculation be checked rather than taken on trust: the threshold is the median plus eight times the median absolute deviation, held above a minimum noise floor of 0.0005 and capped so that the loudest 0.5% of activity in the file is always detected. The dashed line on each histogram marks where the threshold fell.

ID	Threshold	Rule applied	Median	MAD	99.5th pct	Frames scored	Score distribution
F01	0.0005	noise floor	0	0	0.01323014	1000
F02	0.0005	noise floor	0	0	0.01323014	1000

Event timeline

All files merged into a single sequence ordered by wall-clock time, so activity across multiple cameras reads as one chronology. Motion entries mark sustained frame-to-frame change. Lighting entries mark a sustained change in overall brightness, lights switching on, headlights, or an exposure shift, which the motion measure does not register.

#	Wall clock (UTC)	Type	Source file	Offset in file	Duration	What the picture shows
1	2026-08-11 09:15:08.000	motion	F01 front-door_2026-08-11_09-15-00.mp4	00:00:08.000 – 00:00:11.040	3.04s	Something tall and narrow moved across the middle and the right side, from left to right, briskly, for about 3 seconds. A tall and narrow shape about 22% of the frame high, in the centre of frame and the middle right, moving left to right briskly, over about 3% of the picture. peak score 0.022157
2	2026-08-11 09:15:19.000	motion	F01 front-door_2026-08-11_09-15-00.mp4	00:00:19.000 – 00:00:22.040	3.04s	Something tall and narrow moved across the left side and the middle, from right to left, briskly, for about 3 seconds. A tall and narrow shape about 22% of the frame high, in the middle left and the centre of frame, moving right to left briskly, over about 3% of the picture. peak score 0.022126
3	2026-08-11 09:15:28.000	motion	F02 loading-bay_2026-08-11_09-15-20.mp4	00:00:08.000 – 00:00:11.040	3.04s	Something tall and narrow moved across the middle and the right side, from left to right, briskly, for about 3 seconds. A tall and narrow shape about 22% of the frame high, in the centre of frame and the middle right, moving left to right briskly, over about 3% of the picture. peak score 0.022157
4	2026-08-11 09:15:31.000	motion	F01 front-door_2026-08-11_09-15-00.mp4	00:00:31.000 – 00:00:34.040	3.04s	Something tall and narrow moved across the bottom and the bottom right, from left to right, briskly, for about 3 seconds. A tall and narrow shape about 22% of the frame high, in the lower centre and the lower right, moving left to right briskly, over about 3% of the picture. peak score 0.022116
5	2026-08-11 09:15:39.000	motion	F02 loading-bay_2026-08-11_09-15-20.mp4	00:00:19.000 – 00:00:22.040	3.04s	Something tall and narrow moved across the left side and the middle, from right to left, briskly, for about 3 seconds. A tall and narrow shape about 22% of the frame high, in the middle left and the centre of frame, moving right to left briskly, over about 3% of the picture. peak score 0.022126
6	2026-08-11 09:15:51.000	motion	F02 loading-bay_2026-08-11_09-15-20.mp4	00:00:31.000 – 00:00:34.040	3.04s	Something tall and narrow moved across the bottom and the bottom right, from left to right, briskly, for about 3 seconds. A tall and narrow shape about 22% of the frame high, in the lower centre and the lower right, moving left to right briskly, over about 3% of the picture. peak score 0.022116

Enhanced stills

Enhancement here is conventional image processing only: upscaling, sharpening and contrast. No generative or AI-invented pixels are used at any stage. Nothing in an enhanced still is information that was not already present in the recorded frame; the processing only makes existing detail easier to see.

Original still, event-002-F01 — event-002-F01

Enhanced still, event-002-F01 — event-002-F01

Limitations of this analysis

Detection is based on measured frame-to-frame change and overall brightness. It can miss slow, distant, small, or low-contrast movement, and it can flag changes caused by weather, insects, camera shake, auto-exposure, or compression rather than by a person or vehicle.
The "what the picture shows" column describes the MEASUREMENT and nothing beyond it: the rough proportions of the changed area, where in the frame it sits, how much of the picture it covers, and whether its centre travels in a consistent direction and at what pace. Proportions come from a foreground silhouette measured against a per-pixel median background, on a 64×36 grayscale copy at 6 frames per second, corrected for the source aspect ratio. Checked against full-resolution ground truth, the height-to-width figure runs about 9% low, so it is deliberately coarse and the wording carries margin for it.
Where an entry names something, a person, an animal, a vehicle, that name comes from an object detector run locally over the frames of that event, and it is reported only when the object was present in at least half of them. A name is a CLASS, never an IDENTITY. The detector recognises no faces, matches nobody against anything, cannot tell two people apart, and cannot say what anyone was doing or intended. "Someone moved across the yard" means a person-shaped object was measured in those frames and nothing more than that. Detectors also make mistakes in both directions: they miss things that are present, especially when small, distant, dark or partly hidden, and they occasionally name something that is not there. Where no name appears, the entry says "something" and describes only the shape.
An absence of events means no change was measured above the stated threshold. It is not evidence that nothing happened.
Wall-clock times are only as accurate as the source identified in section 2 and as the recorder's own clock, which this process cannot audit or correct.
Enhancement adds no information that was not in the source pixels and cannot recover detail the camera did not capture.
This is a technical processing record. It identifies no person, states no opinion about intent, cause or fault, and is not an investigative finding or expert testimony.
Source footage is retained for 30 days after delivery and then deleted unless the client directs otherwise in writing.