Project
# | Title | Team Members | TA | Documents | Sponsor |
---|---|---|---|---|---|
27 | An Automatic Pet Door(seeking for approval) |
Haijian Wang Haoran Zheng Zhihao Xu |
Yixuan Wang | design_document1.pdf final_paper1.pdf other1.pptx photo2.png photo3.png proposal1.pdf video1.mp4 video |
|
An Automatic Pet Door Team Members: - Student 1 (netid) Haijian Wang (haijian4) - Student 2 (netid) Haoran Zheng (haoranz8) - Student 3 (netid)Zhihao Xu (zhihaox4) 1. Problem For those people living near small natural ecosystems, some small-sized wildlife animals like racoons or lizards may enter their house through pet doors from time to time. If we can design an electronic device attached on the pet door with proper sensors that can distinguish cats and dogs from non-pet animals, then when the pets attempt to enter or exit the house, the pet door will automatically unlock with the help from some external mechanical devices, but if wildlife animals try to enter, the pet door will stay locked. Additionally, the practical use of such device is not limited to pets-scenario, and any problem involved in automatically distinguishing different types of objects and taking different actions can utilize this device because the training sets can be altered to fit different scenarios. 2. Solution and Design Graph The solution to our problem is to design an automatic pet door. We will have several subsystems. The most important subsystem would be a camera module to help us identify the animal at the door. The camera will be connected to an FPGA that runs pre-trained AI models. We will also have weight sensors and motion sensors to further verify that we have the correct type of animal. We will use batteries for our power subsystem and motors to unlock the latches. We would also have a notification subsystem that uses LEDs to indicate the status of the lock and sends out text notifications to users. Our customized PCB will connect every subsystem together and a microcontroller will control everything 3. Solution Components 3.1 Subsystem 1: Camera module and FPGA with pretrained AI implemented An AI programmed on the FPGA board will be trained and tested on recognizing pets' facial images with numerous photos as training, development, and testing sets. After completing the training process and reaching a desirable successful rate, the FPGA would be connected to the camera through PCB. The camera will monitor the outside of the door and send image data to FPGA, so the AI would determine whether the object is pet and generate different signals accordingly. 3.2 Subsystem 2: Assisting sensors 3.2.1 Infrared Motion Sensor: Detect whether some objects are near the pet door, if there is currently no object, then the camera, display device, and FPGA will remain shut down to avoid wasting energy. 3.2.2 Weight Sensor: This sensor serves as a fail-safe, and if the measured weight is lower or higher than the boundary of the expected weight range of normal cats and dogs, then the latch will always be locked even if the camera falsely recognizes the object as a pet. 3.3 Subsystem 3: Microcontroller on a Customized PCB 3.4 Subsystem 4: Sound Notification and Visual Display Device Basic: Single LED, if the latch unlocks, then the LED will light up. Intermediate: a full LED Array to form a rectangular board, if the latch is unlocked, corresponding individual LEDs in the Array will be lit up and display “unlocked”. Advanced: LCD screen 3.5 Subsystem 5: Power Supply Battery for supporting LED and mechanical controller. Using a voltage converter to supply the electric energy for operating the whole system. 3.6 Subsystem 6: Motor and Mechanical locking/unlocking device keeping a door closed until a release mechanism is activated which is related to our multiple sensors. When every sensor is satisfied, the unlocking mechanism will be activated and the door will be opened and will go back into locking status after pets pass through the door. Otherwise, the door will keep closed. 4. Criterion For Success For our project, we need to achieve this system with great efficiency and accuracy for recognizing the general characteristics of pets. We need to make sure the accuracy of the systems could identify animals without being disturbed by other objects. The door will be closed automatically when there are no pets appearing in front of the camera. The recognition of motion, weight, and graph should be satisfied at the same time to identify the pet as the correct type. Any incorrect recognition will make the door keeping closed. The door should be closed in a short time after pets have passed through the door and presented the status of door through the LEDs. The door should respond to the recognition result in a short time. If the recognition is correct/false, the door should open/close and the status of the LED should be changed. |