ESP32-CAM_MJPEG2SD

Application for ESP32 / ESP32S3 with OV2640 / OV5640 camera to record JPEGs to SD card as AVI files and playback to browser as an MJPEG stream. The AVI format allows recordings to replay at correct frame rate on media players. If a microphone is installed then a WAV file is also created and stored in the AVI file.
The application supports:

• Motion detection by camera or PIR / radar sensor
• Time lapse recording
• Audio Recording from I2S or PDM microphones
• Camera pan / tilt servos and lamp control
• Telemetry Recording during camera recording.
• Remote Control of camera mounted vehicle.
• Alert notification using Telegram or Email
• Concurrent streaming to web browser and remote NVR
• Transfer recordings using FTP, HTTPS, WebDAV, or download from browser
• MQTT control.
• Support for peripherals: SG90 servos, MX1508 H-bridge, HW-504 joystick, BMP280, MPU9250, WS2812 Led
• Interface for Machine Learning support.
• Camera Hub feature to access other ESP32-CAM_MJPEG2SD devices.

The ESP32 cannot support all of the features as it will run out of heap space. For better functionality and performance, use one of the new ESP32S3 camera boards, eg Freenove ESP32S3 Cam, ESP32S3 XIAO Sense.

This is a complex app and some users are raising issues when the app reports an warning, but this is the app notifying the user that there is an problem with their setup, which only the user can fix. Be aware that some clone boards have different specs to the original, eg PSRAM size. Please only raise issues for actual bugs (ERR messages, unhandled library error or crash), or to suggest an improvement or enhancement. Thanks.

Changes in version 9.7:

• WebDAV server to enable access to SD card content from client, eg Windows file explorer

Purpose

The application enables video capture of motion detection or timelapse recording. Examples include security cameras, wildlife monitoring, rocket flight monitoring, FPV vehicle control. This instructable by Max Imagination shows how to build a WiFi Security Camera using an earlier version of this code, plus a later video on how to install and use the app.

Saving a set of JPEGs as a single file is faster than as individual files and is easier to manage, particularly for small image sizes. Actual rate depends on quality and size of SD card and complexity and quality of images. A no-name 4GB SDHC labelled as Class 6 was 3 times slower than a genuine Sandisk 4GB SDHC Class 2. The following recording rates were achieved on a freshly formatted Sandisk 4GB SDHC Class 2 on a AI Thinker OV2640 board, set to maximum JPEG quality and highest clock rate.

Frame Size	OV2640 camera max fps	mjpeg2sd max fps	Detection time ms
96X96	50	45	15
QQVGA	50	45	20
QCIF	50	45	30
HQVGA	50	45	40
240X240	50	45	55
QVGA	50	40	70
CIF	50	40	110
HVGA	50	40	130
VGA	25	20	80
SVGA	25	20	120
XGA	12.5	5	180
HD	12.5	5	220
SXGA	12.5	5	300
UXGA	12.5	5	450

The ESP32S3 (using Freenove ESP32S3 Cam board hosting ESP32S3 N8R8 module) runs the app about double the speed of the ESP32 mainly due to much faster PSRAM. It can record at the maximum OV2640 frame rates including audio for all frame sizes except UXGA (max 10fps).

Design

The application was originally based on the Arduino CameraWebServer example but is now independently coded, including contributions made by @gemi254.

The ESP32 Cam module has 4MB of PSRAM (8MB on ESP32S3) which is used to buffer the camera frames and the construction of the AVI file to minimise the number of SD file writes, and optimise the writes by aligning them with the SD card sector size. For playback the AVI is read from SD into a multiple sector sized buffer, and sent to the browser as timed individual frames. The SD card is used in MMC 1 line mode, as this is practically as fast as MMC 4 line mode and frees up pin 4 (connected to onboard Lamp), and pin 12 which can be used for eg a PIR.

The AVI files are named using a date time format YYYYMMDD_HHMMSS with added frame size, FPS recording rate, duration in secs, eg 20200130_201015_VGA_15_60.avi, and stored in a per day folder YYYYMMDD. If audio is included the filename ends with _S. If telemetry is available the filename ends with _M.
The ESP32 time is set from an NTP server or connected browser client.

Installation

Download github files into the Arduino IDE sketch folder, removing -master from the application folder name. Select the required ESP-CAM board using CAMERA_MODEL_ in appGlobals.h unless using the one of the defaults:

• ESP32 Cam board - CAMERA_MODEL_AI_THINKER
• Freenove ESP32S3 Cam board - CAMERA_MODEL_ESP32S3_EYE

Select the ESP32 or ESP32S3 Dev Module board and compile with PSRAM enabled and the following Partition scheme:

• ESP32 - Minimal SPIFFS (...)
• ESP32S3 - 8M with spiffs (...)

NOTE: If you get compilation errors you need to update your arduino-esp32 core library in the IDE to latest v2.0.14 (but not yet v3.0) using Boards Manager

NOTE: If you get error: Startup Failure: Check SD card inserted, or Camera init error 0x105 it is usually a camera board selection issue

NOTE: If you get error: Camera init error 0xffffffff, it is due to some cam boards being sold with only 2MB PSRAM which is insufficient for this app. Warning added to v9.6.

On first installation, the application will start in wifi AP mode - connect to SSID: ESP-CAM_MJPEG_..., to allow router and password details to be entered via the web page on 192.168.4.1. The configuration data file (except passwords) is automatically created, and the application web pages automatically downloaded from GitHub to the SD card /data folder when an internet connection is available.

Subsequent updates to the application, or to the /data folder files, can be made using the OTA Upload tab. The /data folder can also be reloaded from GitHub using the Reload /data button on the Edit Config tab, or by using a WebDAV client.

Browser functions only tested on Chrome.

Main Function

A recording is generated either by the camera itself detecting motion, or by holding a given pin high (kept low by internal pulldown when released), eg by using an active high motion sensor such as PIR or RCWL-0516 microwave radar. In addition a recording can be requested manually using the Start Recording button on the web page.

To play back a recording, select the file using Playback & File Transfers sidebar button to select the day folder then the required AVI file. After selecting the AVI file, press Start Playback button to playback the recording. The Start Stream button shows a live video only feed from the camera.

Recordings can then be uploaded to an FTP or HTTPS server or downloaded to the browser for playback on a media application, eg VLC.

A time lapse feature is also available which can run in parallel with motion capture. Time lapse files have the format 20200130_201015_VGA_15_60_T.avi

Other Functions and Configuration

The operation of the application can be modified dynamically as below, by using the main web page, which should mostly be self explanatory.

Connections:

• The FTP / HTTPS, Wifi, SMTP, and time zone parameters can be defined in Access Settings sidebar button.
  • for Time Zone use dropdown, or paste in values from second column here
• To make the changes persistent, press the Save button
• mdns name services in order to use http://[Host Name] instead of ip address.

To change the recording parameters:

• Resolution is the pixel size of each frame
• Frame Rate is the required frames per second
• Quality is the level of JPEG compression which affects image size.

SD storage management:

• Folders or files within folders can be deleted by selecting the required file or folder from the drop down list then pressing the Delete button and confirming.
• Folders or files within folders can be uploaded to a remote server via FTP / HTTPS by selecting the required file or folder from the drop down list then pressing the File Upload button. Can be uploaded in AVI format.
• Download selected AVI file from SD card to browser using Download button.
• Delete, or upload and delete oldest folder when card free space is running out.

View application log via web page, displayed using Show Log tab:

• Select log type for display:
  • RTC RAM: Cyclic 7KB log saved in RTC RAM (default)
  • Websocket: log is dynamically output via websocket
  • SD card: Unlimited size log saved to SD card
• Use slider to enable SD logging, but can slow recording rate
• Use buttons to refresh or clear selected log

Configuration Web Page

More configuration details accessed via Edit Config tab, which displays further buttons:

Wifi: Additional WiFi and webserver settings.

Motion: See Motion detection by Camera section.

Peripherals eg:

• Select if a PIR or radar sensor is to be used (which can also be used in parallel with camera motion detection).
• Auto switch the lamp on for nightime PIR detection.
• Control pan / tilt cradle for camera.
• Connect a PDM or I2S microphone.
• Connect a DS18B20 temperature sensor.
• Monitor voltage of battery supply on ADC pin.
• Wakeup on LDR after deep sleep at night.

Note that there are not enough free pins on the ESP32 camera module to allow all external sensors to be used. Pins that can be used (with some limitations) are: 3, 4, 12, 13, 33.

• pin 3: Labelled U0R. Only use as input pin, as also used for flashing.
• pin 4: Also used for onboard lamp. Lamp can be disabled by removing its current limiting resistor.
• pin 12: Only use as output pin.
• pin 13: Is weakly pulled high.
• pin 33: Used by onboard red LED. Not broken out, but can repurpose the otherwise pointless VCC pin by removing its adjacent resistor marked 3V3, and the red LED current limiting resistor, then running a wire between the VCC pin and the red LED resistor solder tab.

Can also use the ESP32-IO_Extender repository.

The ESP32S3 Freenove board can support all of the above peripherals with its spare pins.
The ESP32S3 XIAO Sense board has fewer free pins but more than the ESP32.

On-board LEDs:

• ESP32: Lamp 4, signal 33.
• ESP32S3:
  • Freenove: Lamp 48, signal 2.
  • XIAO: Lamp n/a, signal 21.

Other: SD, email, telegram, etc management.

When a feature is enable or disabled, the Save button should be used to persist the change, and the ESP should be rebooted using Reboot ESP button.

Motion detection by Camera

An AVI recording can be generated by the camera itself detecting motion using the motionDetect.cpp file.
JPEG images of any size are retrieved from the camera and 1 in N images are sampled on the fly for movement by decoding them to very small grayscale bitmap images which are compared to the previous sample. The small sizes provide smoothing to remove artefacts and reduce processing time.

For movement detection a high sample rate of 1 in 2 is used. When movement has been detected, the rate for checking for movement stop is reduced to 1 in 10 so that the JPEGs can be captured with only a small overhead. The Detection time ms table shows typical time in millis to decode and analyse a frame retrieved from the OV2640 camera.

Motion detection by camera is enabled by default, to disable click off Enable motion detect in Motion Detect & Recording sidebar button.

Additional options are provided on the camera index page, where:

• Motion Sensitivity sets a threshold for movement detection, higher is more sensitive.
• Show Motion if enabled and the Start Stream button pressed, shows images of how movement is detected for calibration purposes. Grayscale images are displayed with red pixels showing movement.
• Min Frames is the minimum number of frames to be captured or the file is deleted.

Audio Recording

An I2S microphone eg INMP441 is supported by both ESP32 and ESP32S3. A PDM microphone eg MP34DT01 is only supported on ESP32S3. Audio recording works fine on ESP32S3 but is not viable on ESP32 as it significantly slows down the frame rate.

The audio is formatted as 16 bit single channel PCM with sample rate of 16kHz. An I2S microphone needs 3 free pins, a PDM microphone needs 2 free pins (the I2S SCK pin must be set to -1). Pin values (predefined for XIAO Sense) and Use microphone selector are set under Peripherals button on the configuration web page.

The web page has a slider for Microphone Gain. The higher the value the higher the gain. Selecting 0 cancels the microphone.

OV5640

The OV5640 pinout is compatible with boards designed for the OV2640 but the voltage supply is too high for the internal 1.5V regulator, so the camera overheats unless a heat sink is applied.

For recording purposes the OV5640 should only be used with an ESP32S3 board. Motion detection above FHD framesize does not work due to esp_jpg_decode() decompression error.

Recordable frame rates for the OV5460 highest framesizes on an ESP32S3 are:

Frame Size	FPS
QXSGA	4
WQXGA	5
QXGA	5
QHD	6
FHD	6
P_FHD	6

The OV3660 has not been tested.

MQTT

To enable MQTT, under Edit Config -> Others tab, enter fields:

• Mqtt server ip to connect
• Mqtt topic path prefix
• optionally Mqtt user name and Mqtt user password
• Then set Mqtt enabled

Mqtt will auto connect if configuration is not blank on ping success.

It will send messages e.g. Record On/Off Motion On/Off to the mqtt broker on channel /status.
topic: homeassistant/sensor/ESP-CAM_MJPEG_904CAAF23A08/status -> {"MOTION":"ON", "TIME":"10:07:47.560"}

You can also publish control commands to the /cmd channel in order to control camera.
topic: homeassistant/sensor/ESP-CAM_MJPEG_904CAAF23A08/cmd -> dbgVerbose=1;framesize=7;fps=1

Port Forwarding

To access the app remotely over the internet, set up port forwarding on your router for browser on HTTP port, eg:

On remote device, enter url: your_router_external_ip:10880
To obtain your_router_external_ip value, use eg: https://api.ipify.org
Set a static IP address for your ESP camera device.
For security, Authentication settings should be defined in Access Settings sidebar button.

Telemetry Recording

This feature is better used on an ESP32S3 camera board due to performance and memory limitations on ESP32.

Telemetry such as environmental and motion data (eg from BMP280 and MPU9250 on GY-91 board) can be captured during a camera recording. It is stored in a separate CSV file for presentation in a spreadsheet. The CSV file is named after the corresponding AVI file. A subtitle (SRT) file is also created named after the corresponding AVI file. The CSV and SRT files are uploaded or deleted along with the corresponding AVI file. For downloading, the AVI, CSV and SRT files are bundled into a zip file. If the SRT file is in the same folder as the AVI file, telemetry data subtitles will be displayed by a media player.

The user needs to add the code for the required sensors to the file telemetry.cpp. Contains simple example for the GY-91 board.

To switch on telemetry recording, select the Use telemetry recording option bunder the Peripherals button. The frequency of data collection is set by Telemetry collection interval (secs).

Note: esp-camera library conflict if use Adafruit sensor library.

Telegram Bot

Only enable either Telegram or SMTP email.
Use IDBot to obtain your Chat ID.
Use BotFather to create a Telegram Bot and obtain the Bot Token.
In Edit Config page under Other tab, paste in Telegram chat identifier and Telegram Bot token and select Use Telegram Bot.
You may want to make the bot private.
Note that this feature uses a lot of heap space due to TLS.

The Telegram Bot will now receive motion alerts from the app showing a frame from the recording with a caption containing a command link for the associated recording (max 50MB) which can be downloaded and played.

Remote Control

Provides for remote control of device on which camera is mounted, e.g RC vehicle for FPV etc.
Best used with ESP32-S3 for frame rate and control responsiveness.

To enable, in Edit Config page under Peripherals, select Enable remote control.
This will show an extra config button RC Config.
Pressing the RC Config button will allow pins to be defined for:

• SG90 type steering servo
• H-bridge motor control (tested with MX1508)
• On / off lights
• Further parameters for vehicle control.

The streaming view will now have a red button in the top left. Press this to show / hide overlaid steering and motor controls. Camera view buttons can be used to change to full screen. Tethered vehicles can also be controlled via a HW-504 type joystick. Camera view (and microphone and telemetry if enabled) can be recorded.
Motion detection should be disabled beforehand.

Only use this feature if you are familiar with coding and electronics, and can fix issues yourself

Machine Learning

Machine Learning AI can be used to further discriminate whether to save a recording when motion detection has occurred by classsifying whether the object in the frame is of interest, eg a human, type of animal, vehicle etc.

Only feasible on ESP32S3 due to memory use and built in AI Acceleration support.

Only use this feature if you are familiar with Machine Learning

The interface is designed to work with user models packaged as Arduino libraries by the Edge Impulse AI platform. More details in motionDetect.cpp.
Use 96x96 grayscale or RGB images and train the model with for example the following Transfer learning Neural Network settings:

Camera Hub

This tab enables the web interfaces of other ESP32-CAM_MJPEG2SD camera devices to be accessed. To show this tab, in Edit Config page under Other, select Show Camera Hub tab.

In the tab, enter IP address of another camera and press Add IP button, a screen showing an image from the camera is displayed with its IP address overlayed. Repeat for each camera to be monitored. Click on an image to open the web page for that camera.

Press X icon on image to remove that IP address. Press Delete All button to remove all IP addresses. Press Refresh button to update each screen with the latest image from that camera.

The IP addresses are stored in the browser local storage, not the app itself.

Stream to NVR

This feature is better used on an ESP32S3 camera board due to performance and memory limitations on ESP32.

Streams separate from the web browser are available for capture by a remote NVR. To enable these streams, under Edit Config -> Motion tab, select:

• Enable Video stream on /sustain?video=1 for MJPEG stream
• Enable Audio stream on /sustain?audio=1 for WAV stream (need to setup microphone beforehand).
• Enable Subtitle stream on /sustain?srt=1 for SRT stream (need to setup telemetry beforehand, otherwise just a timestamp will be output).

Then save and reboot.

If multiple streams are enabled they need to be processed by an intermediate tool for synchronisation, eg go2rtc (but which does not handle subtitles yet?). See ESP32-CAM_Audio for go2rtc configuration examples. If a recording occurs during streaming it will take priority and the streams may stutter.

WebDAV

A simple WebDAV server is included. A WebDAV client such as Windows file explorer can be used to access and manage the SD card content. In a folder's address bar enter <ip_address>/webdav, eg 192.168.1.132/webdav
For Android, MacOS, Linux see webDav.cpp file.