Ready-made camera software for the RV1126B: 4K sensor capture, hardware encode, audio, RTSP streaming and a standards-compliant ONVIF interface, integrated directly against Rockchip's RKMPI media pipeline.
The silicon
Rockchip's current 4K vision SoC for intelligent cameras: four Arm application cores, a neural engine, a 12-megapixel ISP and a 4K hardware video pipeline, with two MIPI camera inputs.
On this silicon
The RV1126B has no CameraStack GStreamer dependency; its camera and encoders are reached through Rockchip’s RKMPI SDK, as on the RV1106. CameraStack handles platforms like this with a C producer: a compact native adapter, written in C straight against RKMPI and Rockchip’s AIQ libraries, that runs the vendor media pipeline from sensor to hardware encoder and hands encoded video and audio to the portable core. The Go server is built for the SoC’s 64-bit Arm environment and does not link the vendor SDK.
Control flows the other way too. Because the producer controls the encoders directly, settings changed through ONVIF are applied live: frame rate, bitrate and GOP change mid-stream, a codec change rebuilds only the selected VENC channel, and a main-profile change between 1080p and 4K moves only that branch of the pipeline, while the sensor, ISP, audio, sibling stream and analysis path keep running. Synchronisation-point requests map to on-demand IDR keyframes, and every new reader starts on a complete keyframe. Encoding is demand-driven per profile: each hardware encoder runs, and draws power, only while clients consume its stream, while the resident pipeline keeps snapshots, motion events, audio and object metadata flowing with both video encoders idle. The proven profiles are a main profile at 1920x1080 or 3840x2160 at up to 30 fps and a sub profile at 640x360 at up to 10 fps, each selectable between H.264 High and H.265 Main. Those are the resolutions tested on the reference port; other resolutions are supported and can be configured for a product.
The producer captures the camera’s microphone through ALSA and hands G.711 mu-law audio, 8 kHz mono at 64 kbps, to the core as a second RTP track on the main profile, so a single RTSP session carries video and audio together; the sub profile carries video and metadata only. Audio is exposed through the standard ONVIF objects, an AudioSource, AudioSourceConfiguration and AudioEncoderConfiguration, which a client lists and attaches to a profile through Media1 or Media2, and it travels over the same unicast and multicast transports as the video. RTP timestamps advance by sample count, so audio and video stay aligned over long sessions.
The Analytics service runs ONVIF MotionRegionDetector rules: up to four named polygons, each with its own sensitivity and damping. The RV1126B’s IVS block measures a 30x17 activity grid over a resident 960x540 analysis channel, and CameraStack rasterises arbitrary ONVIF polygons onto it, so motion detection never scans full video frames on the CPU and keeps running while both video encoders are idle.
Object detection runs on the RV1126B’s NPU. The ObjectDetector analytics module advertises the models present on the device: the royalty-free portable PicoDet-S, YOLOX-Nano and PP-YOLOE-S, alongside the vendor example detectors, YOLOv5s and YOLOv10n (if their artifacts have been installed). CameraStack advertises only the models genuinely present and does not redistribute vendor detectors whose rights are not independently established, so a licensable model is available only where it has been supplied for the device. A standard ONVIF client can switch the model, confidence threshold and 1 or 5 fps detection rate live. A model change replaces only the private RKNN context and buffers; the sensor, ISP, both encoders, audio and the daemon stay resident. Detected objects are tracked with stable IDs and published as ONVIF scene metadata, and metadata-only clients can subscribe without waking a paused video encoder.
The Imaging service is wired to the ISP through AIQ: brightness, contrast and colour saturation, sharpness, automatic or manual exposure, automatic or manual white balance with independent red and blue gains, and backlight compensation. ImageTooDark is derived from the ISP’s exposure-normalised environment statistic, sampled once a second, which survives a covered lens under automatic exposure.
JPEG snapshots come straight from the hardware at both profile resolutions, including while the video encoders are paused. Rockchip RGN provides four text OSD slots per profile with position, black or white foreground and background across the full transparency range, and plain or date-and-time text. The same OSD state is composited into the independent JPEG snapshots and survives paused encoders, codec replacement and a main-profile resolution change.
The producer builds against Rockchip’s RKMPI middleware and AIQ tuning libraries. Capture runs through VI, the video input module that drives the sensor and ISP; VPSS scales the resident source into the main, sub and analysis outputs; the main and sub outputs feed VENC 0 and VENC 1, the hardware H.264, H.265 and JPEG encoders; the IVS block computes the motion activity grid in hardware; and RGA performs the colour conversion and resize that feed the object models on the RKNPU2 neural engine. The producer builds on the same shared common core as the RV1106 and CameraStack’s other custom-SDK platforms, so framing, configuration, motion-grid mapping, event formatting, OSD rendering, pause state and the object-model decoders are well-tested portable code, with only small RV1126B-specific customisations on top: the RKMPI topology, ISP controls, RKNPU2 tensor handling and the ALSA audio route. The port is proven with a Sony IMX415 sensor at 3840x2160, 30 fps.
Field note
We brought CameraStack up on RV1126B hardware as a 4K network camera with microphone audio, motion detection and selectable NPU object models, and almost none of the portable core had to change.
Read the port storyBuild on it
The port work is already done, at the level of the RKMPI and AIQ SDK your board already has. Use CameraStack Server as your complete camera stack, or as a drop-in ONVIF interface for a pipeline you already have. Evaluation builds are available to teams building real products.
Talk to us about the RV1126B