# Camera Installation Guideline ## 1. Introduction This guide highlights the essential elements of choosing the right camera, installing it correctly, and configuring it for dependable object recognition. Proper camera installation is crucial, as improper installation can result in increased expenses and compromised recognition quality. Thus, we encourage you to familiarize yourself with this manual prior to embarking on the camera installation process. ## 2. Camera Selection When choosing a camera, we strongly recommend paying attention to the following specifications: | Settings | Requirements | | -------------------- | ------------------------------------------------------------------------------------------------------ | | **Type** | Internet Protocol (IP) | | **Image Resolution** | From 1 Megapixel (HD 720p, 1280×720 pixels, 16:9 aspect ratio). Full HD 1080p is recommended. | | **Frame Rate** | FPS at least 5 frames per second. The recommended value is 25 FPS. | | **Tilt Angles** | Recommended variation is not more than 15⁰ vertically and horizontally | | **Shutter Speed** | No more than 1/100 sec. (for example, a shutter speed of 1/25 sec may cause blurring of moving people) | | **Authentication** | Basic. Please make sure that it is configured in the IP camera settings for the RTSP stream. | {% hint style="warning" %} **IMPORTANT !** For optimal performance in video analytics, it is imperative to ensure uniformity in the settings of all video sources connected to our device. This includes maintaining consistency in video stream quality, frame rate, and aspect ratio. Discrepancies in these parameters among different video streams can lead to sub-optimal detection quality and may disrupt or break the detection pipeline itself, as all streams are processed through a singular video analytics pipeline. {% endhint %} This section describes recommendations for installing cameras connected to DEKI Safe to ensure reliable AI performance. #### Installation Example: If a camera is installed at a height of minimum 3 meters with a tilt of 45-degree angle, objects at least one meter away and higher than 170cm can be recognized. Detection is possible until more than two-thirds of the object is visible up to a distance of 5.6m from the camera. For reliable detection of, it is recommended to install cameras at an angle of 45 degrees or larger. Additionally, even though it is recommended to not use angles sharper than 45 degrees, different objects of interest may have different maximum angle requirements.

Note: False or failed detection could result if cameras are not properly installed according to the recommended requirements.

#### The following conditions can reduce the cameras’ performance: * When installing cameras, it is recommended to keep a straight horizon line for aligned images. * Objects should be right-side-up for reliable performance. * It is recommended to not install camera at an sharp angle (Direct under, different object types handle sharper angles differently) * Challenging lighting conditions with shadows or low-light environments * Scenes with strong background lighting such as direct sunlight, backlighting, or direct lighting * Motion blur caused by slow shutter speed or low frame rate * Objects that have similar illumination or color as the background (not enough contrast between the object and the background).

Recommended horizontal field of view

Undesirable condition - when camera gets affected by direct sunlight, backlight or other light.

Undesirable condition - when random objects occlude objects of interest

### 4. Recommended requirements for security monitoring Detection models for Security Monitoring solutions of DEKI Safe can detect persons, vehicles. The number of objects that can be detected simultaneously is 150 in a single frame (if multiple video channels are connected, the maximum simultaneous number of objects is split across them). This section illustrates recommended requirements for detecting each object type. The requirements can vary for each object. Even if all the requirements are met, the performance may vary according to different operating environments. If you want to learn about recommended requirements for detection of objects types that are outside of our security monitoring solutions, please contact: ****. | Object | Recommended speed | Recommended Max. speed | | ----------- | --------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------- | | **Person** | 1m\~1.5m/sec (Cannot guarantee accuracy for objects walking faster than the normal walking speed) | 1m\~1.5m/sec (Cannot guarantee accuracy for objects walking faster than the normal walking speed) | | **Vehicle** |

Below 80km/h (Installed at 45 degrees or sharper)


Below 40km/h (Installed at 35 degrees or sharper)

|

Below 80km/h (Installed at 45 degrees or sharper)


Below 40km/h (Installed at 35 degrees or sharper)

| #### Situations that can result in incorrect detection for all object types: * Object occlusion or partial objects * Random objects that occlude different objects of interest * Objects reflected in reflective surfaces such as mirrors * Objects moving too fast while low frame rate ### 4.1 Recommended requirements for person detection The minimum recommended size supported for detecting a person in an image is 100x100 pixels. The lower the camera height, being closer to the height of the person, the higher the accuracy. #### The following conditions can reduce the accuracy and speed of detection: * Only parts of the body is captured in an image * A person moving faster than a standard walking speed = faster than 1m\~1.5m/sec * In high density crowds (e.g., standing shoulder to shoulder, large crowds) * If a camera is installed at an angle of less than 30 degrees (sharp angles). #### The following conditions can impair detection: * Detection might not be possible if a full body (face/body/leg) is occluded more than 50% from left to right. * Detection might not be possible if the upper or lower body is occluded more than 50%.
### 4.2 Recommended requirements for vehicle detection For vehicles, the minimum recommended size supported for detection is 100 x 100 pixels. The less sharp the camera angle, being closer to the horizontal / frontal view, the higher the accuracy #### **The following conditions can reduce the accuracy and speed of detection:** * Only parts of a vehicle is captured in an image * If a camera is installed at an angle of less than 45 degrees. * If a vehicle is moving faster than 80km/h * If a camera is installed at an angle of less than 35 degrees or a vehicle is moving faster than 40Km/h * In high density crowds (e.g., traffic congestion, waiting at red light) #### The following conditions can impair detection: * Detection might not be possible if the front of a vehicle is occluded more than 75% from left to right. * Detection might not be possible if the front of a vehicle is occluded more than 50% from top to bottom. * Detection might not be possible if the side of a vehicle is occluded more than 25% from left to right. * Detection might not be possible if the side of a vehicle is occluded more than 50% from top to bottom. * Detection might not be possible if a vehicle is overturned.
--- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.deki.ai/deki-safe/installation-manual/camera-installation-guideline.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.