Landing : Athabascau University
  • Blogs
  • Project Proposal, Portable Wildlife Detector for Backcountry Camping

Project Proposal, Portable Wildlife Detector for Backcountry Camping

I want to create a directional, and portable wildlife detection system using a Raspberry Pi 4 that will use an IR filtered camera and IR illuminators to detect the infrared reflection of an animals eyes, and sound an alarm when triggered. The purpose of this device would be intended for camping safety, where bears or other animals may wander into camp looking for food.

 

My idea is to use IR illuminators mounted with an IR camera to create a reflection in the tapetum lucidum of an animals eyes, and detect that reflection in the camera using a machine learning model like TensorFlow (with which I have no experience) running on a Raspberry Pi 4 (which I already own). To increase the sensitivity of the camera, I may be able to use either physical or software IR filtering so that the camera only sees IR light, and amplifies the reflection of the animals eyes. This project is intended to be software development heavy, and does not pose much of a challenge in terms of hardware requirements. I may add an ultrasonic chirp to the unit to try and get the attention of animals nearby so that they look at the camera, however with this sensor being deployed in a campsite this may not be needed.

 

I originally wanted to create some sort of perimiter alarm system using Panasonic PIR sensors with a 50' range, attached to ESP32 WiFI development boards, which were connected to a central processing/alert unit in the camp (I will also include the proposal I hade made for that project). However after researching the use of PIR sensors, microwave sensors, mmwave, ultrasonic, and other solutions for a perimiter alarm, this new idea seems to address the issues of false positives when using other types of sensors. Focusing on the reflection of the tapetum lucidum also avoids false positives from humans, as well as detects large and small animals. Using object detection would likely fail also in this scenario, where animals are partially obscured behind trees and such.

 

Components needed:

 

  • Raspberry Pi 4 8GB (already owned)
  • Compatible battery pack (I own many LiPo quadcopter batteries I might be able to adapt to work)
  • Compatible Camera with IR coating removed such as the RasPi NoIR camera
  • Powerful IR spotlight illuminators (likley buy premade, as a battery powered flashlight for proof of concept)
  • Speaker for Raspberry Pi for Chirp in the ultrasonic range, outside human hearing but inside animal hearing.
  • LCD display for Raspberry Pi (possibly full color for aiming)
  • 3D printed weather resistant case if time permits (im a 3d printing nerd)

 

Software:

  • Tensor flow and custom trained model for detecting bright points on video.

 

Im not sure if I need approval or not before I start this project, as it is quite software heavy as opposed to implementing the principles learned in the course. However I could not find a good hardware heavy solution to this wildlife perimeter alarm solution. Note that this project will take considerable software development time, as well as creating custom training data and testing on my cats (haha). I even have a backpacking trip planned near the end of my term, and if all goes well would like to make a fun video testing the system.

 

Let me know if this is adequate for a project as I have likely purchased components already (less than 2 months left in my term).

Original proposal that I dont think would work due to the inability for a PIR sensor to differentiate between an animal and environmental motion:

I would like to build an outdoor perimeter alarm system to detect wildlife in a perimeter around a central control unit. The system will be designed to be carried into backcountry campsites, and setup quickly using aids from the devices. The system will feature 3-4 sensor units with PIR sensors for detecting infrared motion up to 50' from the sensor unit, and a central control unit connected to the sensor units through WiFi, with status monitoring for the sensors, signal processing, alert system, LCD display, and possibly more features depending on time. The system will use a P2P or Ad-hoc wifi network to connect sensor units and central control unit over WiFi, as well as use RSSI (received signal strength indication) to determine the rough distance between each sensor and the central control unit for setup purposes.

 

The system would be deployed from camp, with the user first setting the main control unit in place in camp. Then a sensor unit will be walked out to a location at x distance (TBD) from the main control unit (depending on sensor count). The LCD display on the unit will show the user the distance from the main control unit, aiding in setup. Once the first sensor is deployed, the second sensor will show distance from the main control unit as well as the first sensor, and so on for all sensors, allowing users to set up an effective perimeter with guaranteed coverage around their campsite. The distance of sensors from main control unit will likley be about 40 feet, however the sensors will overlap an area between them as well as project out away from the main control unit, increasing the effective range.

 

Components list:

Sensor units (each):

  • 1 only Raspberry Pi Pico W board (264 kb RAM, dual core processor, 2.4 ghz WiFi)
  • alternative is Pi Zero W for external antenna support.

2 only EKMB4311111K Panasonic Passive infrared sensors (digital output signal)Possibly Raspi Pico compatible LCD display, and contrast adjustment potentiometer if necessary
Compatible Battery
weather resistant 3d printed case out of ASA (weather resistant ABS) (designed by me)

 

Central Camp Unit

  • Rasberry PI 4 8GB (overkill but already own one)
  • Speaker or buzzer for alarm
  • RGBLED to show status of system (from SparkFun kit)
  • LCD display for status of individual sensors (likley the one included in the SparkFun kit)
  • Compatible battery
  • Weather resistant 3d printed case (possibly DIY design, however may need to pull one from the internet to save some time)

Comments