AI-Powered Smart Security Camera Platform Development

Adequate Infosoft designed and developed an AI-powered mobile application, an intelligent IP camera management platform available on both Android and iOS for AIS (Advanced Info Service). This project showcases Adequate Infosoft's expertise in cross-platform mobile application development, IoT system integration, and AI-driven security solutions.

Built entirely with .NET MAUI, app unifies custom AI camera hardware and multiple third-party AI camera SDKs including XMEye and JFTech under a single, seamless mobile experience.

The platform delivers real-time intelligent alerts for fall detection, face recognition, fire and smoke detection, pet monitoring, and elderly care going far beyond what traditional CCTV or IP camera apps can offer.

Via the Microsoft Azure IoT architecture, the MQTT-based edge AI camera communication as well as the multi-vendor SDK integrations, this case study illustrates how Adequate Infosoft is building enterprise-level, AI-enabled mobile agreements that are ready for production and readily scalable for commercial use.

Technology:

.NET MAUI | ASP.NET Core | Third-Party AI Camera SDKs | Cloud Backend

AICAM App is a next-generation AI-powered IP camera management platform developed to deliver intelligent, real-time surveillance and monitoring across multiple use cases — from home security and elderly care to pet monitoring and child safety.

Unlike conventional CCTV systems that simply record motion, understands what is happening in the camera's field of view and generates context-aware alerts. The system was architected to support both our proprietary custom AI camera hardware and third-party AI camera devices from multiple manufacturers, all unified under a single cross-platform mobile application built with .NET MAUI.

The security camera market has many basic motion-detecting devices but lacks any intelligent capability so consumers and businesses need a platform that can:

• Identify a person, vehicle, or pet as well as identify the individual, not just all motion events.
• Provide event-based notifications (such as fall detection, smoke/fire detection, eating behavior detection, unknown person detection) versus simply creating multiple clips that all show motion.
• Provide native support for both iOS and Android devices from one codebase.
• Allow for integration of various camera types and brands into the same application.
• Allow users to view and analyze the history of events, rather than requiring them to review hours of video for specific events.

The ultimate goal was to develop a mobile application that will act as an intelligent and expandable software solution to integrate various camera hardware types, brands, and manufacturers to provide AI inference from the variety of camera hardware solutions.

The platform is composed of four major layers:

1. Cross-Platform Mobile Application (.NET MAUI)

2. Cloud Backend (ASP.NET Core + REST/WebSocket APIs)

3. Custom AI Camera Hardware & Firmware

4. Third-Party AI Camera SDK Integrations

Our mobile + desktop app was created using the .NET MAUI (Multi-platform App UI) platform from Microsoft. .NET MAUI is a cross platform framework that compiles a single codebases written in C# to create native applications for both Android and iOS.

Choosing this method of creation was a strategic decision which allowed us to reduce the time spent developing this application by 40% when compared to creating two separate native codebases for both platforms while retaining the benefits of developing a native user interface on each of these platforms.

Key technical aspects of the MAUI implementation:

Live Video Streaming was accomplished by using platform specific media renderers through MAUI's DependencyService model. iOS utilized AVFoundation for low latency RTSP/RTMP stream playback and Android utilized ExoPlayer through MAUI's native interop layer.

Push Notifications were implemented using Firebase Cloud Messaging (FCM) on Android and Apple Push Notification Service (APNs) on iOS. Both notifications ultimately go through a shared MAUI notification service interface.

Real-Time Alerts are sent to the mobile app using a .NET SignalR client (shared) and WebSockets to push live alerts for events such as fall detection, fire detection and unknown people from the backend to the mobile app with sub-second latency.

Offline Resilience using an in-device SQLite database (using Entity Framework Core) to cache device settings, event history metadata and alert preferences gives the app the ability to function without connectivity

Authentication for logging into the AIS mobile application was done through mobile phone number with OTP verification and was backed by JWT authentication from the backend API.

A Multi-Camera Dashboard whereby the app can display and manage multiple camera feeds at the same time as well as display thumbnail camera streams in a grid layout using the MAUI CollectionView with custom cell templates has been created.

The backend of our Web API was created using ASP.NET Core and is hosted on cloud-based infrastructure with the following:

Device Management API:

The Device Management API provides RESTful endpoints for registering, configuring, and managing (getting and deleting) camera devices for each user account. Each camera device is associated with an AIS subscription number (to allow for carrier-billing as part of a service package).

AI Event Processing Pipeline:

The AI Event Processing Pipeline is a separate microservice that receives AI inference results from either the camera firmware or from third-party SDK (software development kit) callbacks and processes these results into structured AI event objects (event type, timestamp, confidence score, event thumbnail snapshot).

Once the AI Event Processing Pipeline creates the event (including this data), it saves this data to a relational database and pushes it out to the relevant mobile clients (via SignalR hubs).

Video Storage & Retrieval:

The video clips captured with an event tag are stored in cloud blob storage (as opposed to file system storage).

The backend creates an indexed event timeline for each of the devices, which adds the capability of the AI Playback feature in the app. With the AI Playback feature, users can simply filter or view the captured footage (for example, to show all times today that a person was eating) without manually scrubbing through the footage.

Daily Highlights Generation:

The Daily Highlights are generated nightly as described above via a background service of background job processing, hosted as an ASP.NET Background Service, which combines together the key events (AI-determined) detected by each camera during the day into short Daily Highlight Reels so that these can then be displayed as such in the app.

Statistics API:

Each day, the stats regarding each of the events generated by all the cameras are aggregated and exposed to the app’s dashboard UI through a statistics endpoint (e.g., “The pet was detected 12 times today; unknown visitor at 3:47 PM”).

The platform's proprietary AI cameras run embedded firmware with an on-device inference engine.

The cameras perform edge AI processing , running lightweight object detection and behavior recognition models directly on the camera's SoC (System-on-Chip) before transmitting only the event metadata and clip thumbnail to the backend. This dramatically reduces bandwidth consumption and cloud processing costs.

The back end communicates with the custom camera via MQTT (Message Queuing Telemetry Transport), a lightweight pub/sub protocol developed specifically for IoT devices. Each camera also publishes event payloads to an MQTT topic on the MQTT broker so that the back end can subscribe to them via a message broker.

The architecture allows for five configurable AI feature profiles per device:

• AI Smart Outdoor Security : vehicle, person, and animal recognition with entry/exit alerts
• AI Smart Indoor Security : familiar vs. unfamiliar face recognition
• AI Smart Elderly Care : fall detection, diet monitoring, activity tracking
• AI Smart Pet Care : pet behavior recognition and anomaly alerts
• AI Smart Kids Care : homework vs. play detection, security alerts

A core architectural requirement was interoperability with AI cameras from multiple third-party manufacturers. Rather than building hardware lock-in, the platform was designed as an open integration layer.

We integrated SDKs from several third-party AI camera vendors, including the XMEye and JF Software camera ecosystem (as evidenced by the XMEye privacy policy endpoint and JF Software as the publishing developer on Google Play).

Adapters Pattern For Third Party SDK Integrations

The integration of the third-party camera SDKs via the Adapter Pattern by wrapping each of the SDKs with a common ICamera interface from the backend to normalize the vendor-specific data formats used by the vendors for sending events to the platform into a single unified event model within the platform.

Each third-party SDK integration had a number of components that were included to support the integration,

SDK Initialization & Device Pairing:

The vendor SDKs had to be initialized with the serial number of the device, and a unique authentication token from the vendor in order to successfully connect to the SDK. Depending on the vendor, they either required peer-to-peer connection or used cloud relays to communicate with the SDK.

Event Mapping Layer:

The vendor would provide a list of vendor specific event codes (for example, the event code "motion event # 4" from a vendor SDK) that were mapped to the AiCAM's event taxonomy (for example, PersonDetected, VehicleDetected, FallDetected).

Stream Protocol Normalization:

Vendors used multiple streaming protocols (RTSP, HLS, and proprietary P2P) to send their video streams to the backend. The backend had a unified streaming proxy service which normalized all of the streaming protocols into HLS so that they could be consumed by the MAUI app.

Privacy & Data Compliance:

All third-party SDK data flowed through our backend and not directly to the vendor's cloud so that all user data was encrypted across networks and no raw video data was sent to third-party vendor clouds without user consent.

The combined AI capability set across custom and third-party cameras includes:

Image Recognition:

Human detection, familiar/unfamiliar face recognition, pet identification (cat, dog, bird), vehicle recognition, behavioral activity detection (eating, drinking, sleeping, exercising)

Sound Recognition:

Behavioral Analytics:

Fall detection for elderly care, activity pattern monitoring, diet tracking, anomalous behavior flagging

The AiCAM platform successfully delivered a unified AI camera management experience across heterogeneous hardware from multiple vendors.

Using .NET MAUI eliminated the need for parallel iOS and Android development teams. The event-based AI architecture replaced raw video storage with intelligent, searchable event timelines.

The platform was successfully published on both major app stores and commercially deployed through AIS (Advanced Info Service).

Live Production Deployment

With over 10,000 downloads from the Google Play store, this platform uses state-of-the-art AI features to provide an enhanced surveillance experience, including face recognition, fall detection, fire and smoke detection, and pet monitoring, as well as sending users immediate alerts when events occur - which is far superior to conventional CCTV security systems.