Guides · Setup
Shopping for security cameras usually starts with a number: 2MP, 4MP, 8MP. Higher sounds better, and often it is, but the headline figure hides what really matters. What counts is how many pixels land on the thing you care about, whether that is a doorway, a license plate, a person approaching a gate or a package on a porch. AI analysis depends on those pixels. Understanding the idea of pixels on target helps you choose the right camera and place it where it can do its job.
Pixel density is the number of pixels covering each meter (or foot) of the scene at a given distance. A camera with 2,560 horizontal pixels that views a scene 10 meters wide delivers about 256 pixels per meter. Stretch the same camera across a 40 meter yard and the density falls to 64 pixels per meter. The sensor did not change, but the detail available for any one object dropped to a quarter. This is why a wide, distant view with a high-megapixel camera can still look soft when you zoom in on a person.
Security planners often describe tasks in levels. Spotting that something is present needs the least detail. Recognizing what it is, such as a person versus an animal, needs more. Confirming specific details, like a clothing color or an object being carried, needs more again. The commonly cited European standard for CCTV design (EN 62676-4) uses roughly 25, 62, 125 and 250 pixels per meter for detection, observation, recognition and identification respectively. Treat these as planning guides rather than guarantees, since lighting, motion and lens quality all affect results, but they give a useful sense of scale.
For AI analytics, the practical lesson is that detecting a person walking through a frame needs far fewer pixels than reliably telling a delivery driver from a trespasser. The more you expect the system to understand, the more pixels it needs on the subject.
Instead of choosing a camera first, define what each camera must accomplish. A camera covering a wide parking area may only need to detect vehicles and people moving through. A camera at an entrance may need to show enough detail to verify who is at the door. A camera watching a cash counter may need still more. Decide the width of the area, the distance to the subject and the level of detail required, then calculate the pixel density you will get. If it falls short, narrow the field of view, move the camera closer, use a longer lens or add a second camera rather than hoping software will fill the gap. Our guide to where to place security cameras covers positioning in more depth.
Lighting. A high-resolution image in poor light can be noisy and less useful than a lower-resolution image in good light. See our article on AI security cameras at night for what to expect after dark.
Lens and sensor quality. Two cameras with the same megapixel count can differ greatly in sharpness and low-light behavior.
Compression. Heavy compression discards detail, so bitrate settings matter as much as the sensor. The trade-off with storage is explained in our footage retention and storage guide.
Motion. Fast-moving subjects blur if shutter speed is too slow, which reduces effective detail regardless of resolution.
Higher resolution has costs. Files are larger, so storage and bandwidth requirements rise, and weaker networks can struggle. This ties directly to reliability, as discussed in PoE versus Wi-Fi cameras. Analytics may also process a scaled-down version of the image to run smoothly, which means extra resolution is not always used for detection. A well-placed 4MP camera aimed at the right area frequently outperforms a poorly aimed 8MP camera.
Insufficient detail is a quiet cause of false alerts. When a distant subject is only a handful of pixels, the system has less to distinguish a person from a shadow, a branch or an animal. Raising pixel density on the areas that matter, and masking areas that do not, helps the analytics make better decisions. For more on this, read how to reduce false alarms with AI cameras.
Measure the width of the scene at the distance that matters. Divide the camera's horizontal pixel count by that width to estimate pixel density. Compare the result with what you need the camera to do. Check the lighting at the times you care about most. Confirm that your network and storage can handle the bitrate. After installation, review real footage from the camera at the farthest point you need to cover, not only the center of the frame. If you are working with existing equipment, our article on turning cameras into AI security explains how to assess what you already have.
Megapixels are a starting point, not an answer. Think in pixels on target: how much detail reaches the subject you need to detect, recognize or verify. Define each camera's job, calculate the density it will deliver, support it with good lighting and sensible compression, and test it in place. That approach gives AI analytics the clear footage it needs to work reliably.