Sheep Worrying: Getting Evidence That Actually Identifies the Dog and the Owner
Sheep worrying is one of the few rural crimes where the perpetrator usually comes back. Not the dog, necessarily. The owner. The same footpath, the same time of day, often the same week. Which means it is one of the few rural crimes where a camera genuinely has a chance, because you are not trying to catch a one-off raid at three in the morning. You are trying to identify somebody who walks past your ground regularly. The reason so few cases go anywhere is not that nobody saw anything. It is that what got recorded cannot identify a dog, never mind a person. Last updated: November 2026. Why most footage fails Three failures account for nearly all of it. The camera was too far away. A dog at eighty metres is a moving smudge on a 4MP sensor. You can tell something happened. You cannot tell a lurcher from a labrador, and no officer is going to act on a smudge. It triggered too late. Passive infrared motion detection fires on a heat signature crossing the sensor. By the time a camera wakes, focuses and starts writing, the dog is already into the flock and whoever let it off the lead is thirty metres down the path and facing away. It only covered the sheep. Understandable, and the wrong instinct. The sheep tell you that it happened. The gate, the stile and the parking spot tell you who. What to cover instead Think about it as a sequence rather than a scene. Somebody parks, comes through an access point, lets a dog off, and leaves the same way. Four moments, and the two useful ones are at the start and the end. The access point. Stile, kissing gate, field gate, gap in the hedge. This is where a person is closest to a camera, moving slowly, usually facing it on the way in. A camera at five to ten metres on an access point produces an image you can put in front of somebody. A camera at eighty metres across a field does not. The parking spot. The lay-by, the verge, the farm entrance people use. A vehicle and a number plate is the single most actionable thing you can hand over, and it is the one piece of evidence that survives the owner denying it was their dog. Our guide to ANPR on rural gates and entrances covers the practical side of plate capture, including why read angle matters far more than megapixels. The flock itself, as context. Worth having, but it corroborates. It rarely identifies. The detection problem, honestly This is where a lot of farm camera advice goes quiet, so let us be direct about it. Standard motion detection on a field camera is triggered by warm things moving. In a field containing sheep, warm things move constantly. You either set sensitivity high and get hundreds of alerts a day until you stop looking at them, or you set it low and miss the event. Two things genuinely help. AI classification rather than raw motion. A camera that distinguishes a person and a vehicle from general movement will ignore the flock and alert on the walker. That is the difference between a camera you check and a camera you mute. Several of our 4G AI cameras do this detection on the camera itself, which matters because it means the filtering happens before anything is sent anywhere. Putting the trigger where the people are. An access point camera has a narrow, predictable field of view with a path through it. Motion there means a person, almost by definition. You do not need clever classification to make a gate camera useful. Signage, and why it helps you twice Signs do two jobs, and the second is the one people forget. The first is deterrence, which is real but modest. The second is legal: if your camera covers any part of a public right of way, and most access-point cameras do, you need signage saying recording is taking place and who is responsible. That is not optional, and footage from an unsigned camera covering a footpath is evidence somebody can argue about. We covered the detail in cameras and public footpaths, and the short version is that a clear sign at the access point makes your footage harder to challenge and quietly tells the dog walker that this particular field is watched. Both help. What to do in the hour after it happens The footage is only half of it. What you do next decides whether it goes anywhere. Report it, even if you are not hopeful. Livestock worrying is a specific offence and forces record it as such, but only if it is reported. Pattern data is what gets a footpath patrolled. Export the clip, do not just describe it. Pull the file off the camera or platform and keep a copy of your own. Cameras overwrite, platforms expire, and the one clip you needed is the one that rolled off. Note the time precisely. If the dog came through at 14:20, the parking camera at 14:12 and the gate camera at 14:17 become a chain rather than three unrelated clips. Photograph the stock and get the vet out. Worrying that does not break the skin still causes abortion and deaths from stress days later. That is loss, it is claimable, and it needs documenting at the time. Our guide on getting footage the police will act on goes through the handover in more detail. It was written about theft, and the principles are identical. A realistic setup For a holding with two or three regularly worried fields, this is usually enough: One camera per access point people actually use, mounted low enough to catch faces rather than hats, five to ten metres from the stile or gate. One camera covering the parking spot, angled for plates rather than for a wide view. Signage at every access point, naming you as the responsible party. AI person detection if the camera is anywhere it can see the flock, so that you are not fighting constant alerts. Power and signal are usually the practical constraint rather than cost, and both are solvable without a trench. Our solar 4G camera guide covers sizing for a camera that needs to run through a British winter, and whole-farm camera networks without SIM cards is the better route if you have several access points clustered within a kilometre or two of the farmstead. What a…