Thermal Vision

The Thermal Vision integration allows for the use of Thermal Imaging sensors in Home Assistant.

These can be used to detect humans much more reliably than motion sensors as do not require the subject to be moving. They also produce a pretty thermal camera image!

This is a rework of eyalcha/thermal that corrects a number of issues, makes it compatible with HA 2016.6 forward, and adds new functionality.

Differences

This is may not be a complete list:

• The platform/domain name is thermal_vision instead of thermal
• More resilient connections
• Fixes issues:
  • eyalcha/thermal#1
  • eyalcha/thermal#2
  • eyalcha/thermal#5
  • eyalcha/thermal#6
• Works with HA 2016.6 and up
• Allows for both imperial and metric measurements
• Optional auto-ranging capabilities
• Optional in-camera overlay of metrics

Installation

There is three options to install this integration; HACS is the easiest.

INSTALLATION VIA HACS

Just go into the HACS Integrations section, search for the "Thermal Vision" repository, then click Install!

MANUAL INSTALLATION VIA HACS

Follow the instructions at https://hacs.xyz/docs/faq/custom_repositories/ using the URL https://github.com/TheRealWaldo/thermal

MANUAL INSTALLATION

1. Download the zip file from latest release.
2. Unpack the release and copy the custom_components/thermal_vision directory into the custom_components directory of your Home Assistant installation.
3. Configure the thermal_vision sensor and/or camera.
4. Restart Home Assistant.

Configuration

Camera

# Example configuration.yaml entry

camera:
  - platform: thermal_vision
    host: http://192.168.0.10

Main Options

Parameter	Required	Description
platform	Yes	Platform name thermal_vision
name	No	Friendly name Default: Thermal Vision
host	Yes	IP address or hostname of your Thermal sensor server
pixel_sensor	Required if host is not set; sensor containing the base64 encoded pixels
verify_ssl	No	Verify SSL or not Default: false
width	No	Image width in pixels Default: 640
height	No	Image height in pixels Default: 640
preserve_aspect_ratio	No	Preserve aspect ratio (ignores height) Default: true
rotate	No	Rotate image Default: 0
mirror	No	Mirror image true / false Default: false
format	No	Camera image format (jpeg, png) Default: jpeg
min_temp	No	Min temperature Default: 26
max_temp	No	Max temperature Default: 32
auto_range	No	Rather than use a static minimum and maximum temperature, auto adjust based on the content Default: false
min_diff	No	Minimum difference when auto-ranging. Favors cold. Default: 4
sensor	No	Sensor related configurations (see below)
interpolate	No	Interpolation related configurations (see below)
cold_color	No	Cold color Default: indigo
hot_color	No	Hot color Default: red
session_timeout	No	Timeout in seconds for polling Thermal sensor server Default: 2
overlay	No	Add an overlay to the image to visualize min/max temperature Default: false

Interpolate

Parameter	Required	Description
method	No	Interpolation method (bicubic, linear, disabled) Default: bicubic
rows	No	Number of rows in interpolated data Default: 32
cols	No	Number of columns of interpolated data Default: 32

Sensor

Parameter	Required	Description
rows	No	Number of rows in sensor data Default: 8
cols	No	Number of columns in sensor data Default: 8

State and Attributes

Attributes

Attribute	Description
fps	Approximate frames per second based on response time of sensor
min	Minimum value represented by the cold_color in the image
max	Maximum value represented by the hot_color in the image

Sensor

# Example configuration.yaml entry

sensor:
  - platform: thermal_vision
    host: http://192.168.0.10

Main Options

Parameter	Required	Description
platform	Yes	Platform name thermal_vision
name	No	Friendly name Default: Thermal Vision
host	Yes	IP address of your Thermal sensor server
verify_ssl	No	Verify SSL or not Default: false
scan_interval	No	Get raw data interval in seconds Default: 60
sensor	No	Sensor related configurations (see below)
roi	No	Sensor region of interest (see below)
state	No	Sensor state type (average, max, min, sensor_temp, person_detected (deprecated, use binary_sensor)) Default: average

Sensor

Parameter	Required	Description
rows	No	Number of rows in sensor data Default: 8
cols	No	Number of columns in sensor data Default: 8

ROI

Parameter	Required	Description
left	No	Left pixel index [0:cols-1] Default: 0
top	No	Top pixel index [0:rows-1] Default: 0
right	No	Right pixel index [0:cols-1] Default: 7
bottom	No	Bottom pixel index [0:rows-1] Default: 7

State and Attributes

State

The sensor state can be either average, max, min, or person_detected based on state in the configuration. It defaults to average.

Attributes

All values are affected by the ROI configuration.

Attribute	Description
average	Average temperature of all pixels in current capture
max	Maximum temperature of all pixels in current capture
min	Min temperature of all pixels in current capture
min_index	The index where the min temperature was detected (1 Dimensional)
max_index	The index where the max temperature was detected (1 Dimensional)
sensor_temp	The temperature of the sensor itself (if the sensor provides it)
person_detected	A boolean representing whether the sensor detected a person or not. Must use latest firmware! (deprecated, use binary_sensor)

Binary Sensor

# Example configuration.yaml entry

binary_sensor:
  - platform: thermal_vision
    host: http://192.168.0.10

Main Options

Parameter	Required	Description
platform	Yes	Platform name thermal_vision
name	No	Friendly name Default: Thermal Vision
host	Yes	IP address of your Thermal sensor server
verify_ssl	No	Verify SSL or not Default: false
scan_interval	No	Get raw data interval in seconds Default: 60

State and Attributes

State

Returns on when the sensor claims to detect a person, off when it does not. Device class is occupancy.

Sensor Hardware and Firmware

Sensors are based on a simple JSON interface.

Sensor	Firmware
AMG8833	Firmware for a simple sensor using an ESP8266 (I'm using a D1 Mini) can built using TheRealWaldo/esp8266-amg8833. This firmware is still under development! Also, there's an experimental implementation using ESPHome at TheRealWaldo/AMG8833-ESPHOME. You can leverage this by setting pixel_sensor to the sensor created by ESPHome and not using host. Note: this only works with the camera at this time.
MLX90640	Another PlatformIO project using an ESP32 and the MLX90640 cab be found at pixelsquared/thermal_vision-ESP32-MLX90640.

Known Issues

Interpolation only works with sensors that have even dimensions (i.e. 8x8, 64x64, etc.). A work-around is to disable interpolation:

camera:
  - platform: thermal_vision
    host: http://192.168.0.10
    sensor:
      rows: 24
      cols: 32
    interpolate:
      method: disabled