Smart Air Cooling: Implemetation of IoT for face Detection using arduino, OpenCV and haarcascade
Team Members:
- Mohammad Syafiq Bin Enchek Muda (B031910178) (Leader)
- Abdillah Safwan Bin Yusop (B031910120)
- Muhammad Ridhwan Bin Razalee (B031910197)
- Farid Hazeeq Bin Burhanuddin (B031910427)
Objectives: The objective for this project are mainly for
- Reduce amount of workload needed to adjust fan position
- Create a small portable cooling fan just like a mini air-conditioner
- The fan will automatically target person/moving object using a camera.
- Increase efficiency of the power and air-cold as normal fan will only blow in static or sometime it rotate when there is no person standing there.
Currently the temperature in Malaysia now are soaring high as the weather in today world are increasing rapidly from time to time. This become major problem for Malaysia and also to those nation in the earth equator line (Khatulistiwa). Highest recorded temperature in malaysia are 36.4 degree celcius on February 2021, potentially making the coming weeks blistering hot. Moreover, based on the news taken from TheStar and BHarian, MetMalaysia director-general Jailan Simon said the country is currently in the last phase of the northeast monsoon season that will last until mid-March.
During this period, he said, rainfall is expected to be low in Peninsular Malaysia, particularly in the northern region. Those people who working on air-conditioner place will not have much trouble of ii but what happen to the people that dont have air-conditioner at the work place. As we know, heat place may reduce our energy to work efficently, especially during this MCO time due to Covid-19 happen on 2021 now, where most people work at home and which is most poeple dont have air-conditioner like their office.
Due to that, our team decide to make a Smart Air Cooling to solve this problem. Of course, we know that there a lot of product out there selling the air cooling fan. But this is not a normal type air cooler. It is a smart air cooler which user AI technology to automatically blow cool air toward the person. The Idea is to make a fan with servo to automatically targeting the nearest person posibble based on AI movement tracking or Person tracking depending on situation.
This is a small project that combining IoT and AI together to make a small scale air cooling which automatically blow the cold air toward the target user.
Figure 1 shows the static fan blow only on straight where user need to stand in front of it.
For this dataset we will be using haarcascades which is an Object Detection using Haar feature-based cascade classifiers for the effective object detection method proposed by Paul Viola and Michael Jones based on their paper, "Rapid Object Detection using a Boosted Cascade of Simple Features" in 2001. It is a machine learning based approach where a cascade function is trained from a lot of positive and negative images. It is then used to detect objects in other images. Initially, the algorithm needs a lot of positive images (images of faces) and negative images (images without faces) to train the classifier. Then we need to extract features from it. For this, Haar features shown in the below image are used. They are just like our convolutional kernel. Each feature is a single value obtained by subtracting sum of pixels under the white rectangle from sum of pixels under the black rectangle.
For this, we apply each and every feature on all the training images. For each feature, it finds the best threshold which will classify the faces to positive and negative. Obviously, there will be errors or misclassifications. We select the features with minimum error rate, which means they are the features that most accurately classify the face and non-face images. (The process is not as simple as this. Each image is given an equal weight in the beginning. After each classification, weights of misclassified images are increased. Then the same process is done. New error rates are calculated. Also new weights. The process is continued until the required accuracy or error rate is achieved or the required number of features are found).
The final classifier is a weighted sum of these weak classifiers. It is called weak because it alone can't classify the image, but together with others forms a strong classifier. The paper says even 200 features provide detection with 95% accuracy. Their final setup had around 6000 features. (Imagine a reduction from 160000+ features to 6000 features. That is a big gain).
for the source code we will provided link below and also the folder for the haarcascades. Here is the result of running the code above and using as input the video stream of a in webcam:
NOTED: The haarcascades that will be used will be haarcascade_frontalface_default.xml, the user may change the haarcascades file and the code based on that file as they like.
Additional Resources:
- Paul Viola and Michael J. Jones. Robust real-time face detection. International Journal of Computer Vision, 57(2):137–154, 2004. [224]
- Rainer Lienhart and Jochen Maydt. An extended set of haar-like features for rapid object detection. In Image Processing. 2002. Proceedings. 2002 International Conference on, volume 1, pages I–900. IEEE, 2002. [132]
- An interesting interview regarding Face Detection by
- OpenCV Face Detection: By Liz Miller
The following directory is our structure of our project:
- $ tree --dirsfirst
- Smart Air Cooler v1.3
- ├── arduino_code
- │ ├── connection_tester
- │ └── LED tester.ino
- │ ├── cmoving_camera
- │ └── moving_camera.ino
- │ └── static_camera
- │ └── static_camera.ino
- ├── img
- │ └── (Used for README.md).
- ├── haarcascades
- │ ├── haarcascade_eye.xml
- │ ├── haarcascade_eye_tree_eyeglasses.xml
- │ ├── haarcascade_frontalcatface.xml
- │ ├── haarcascade_frontalcatface_extended.xml
- │ ├── haarcascade_frontalface_alt.xml
- │ ├── haarcascade_frontalface_alt_tree.xml
- │ ├── haarcascade_frontalface_alt2.xml
- │ ├── haarcascade_frontalface_default.xml
- │ ├── haarcascade_fullbody.xml
- │ ├── haarcascade_lefteye_2splits.xml
- │ ├── haarcascade_licence_plate_rus_16stages.xml
- │ ├── haarcascade_lowerbody.xml
- │ ├── haarcascade_profileface.xml
- │ ├── haarcascade_righteye_2splits.xml
- │ ├── haarcascade_smile.xml
- │ └── haarcascade_upperbody.xml
- ├── haarcascade_frontalface_default.xml
- ├── Connection_testing.py
- ├── Face_detection_test.py
- └── Smart Air Cooler.py
The dataset/ directory contains the data described in the “Smart Air-Cooler dataset” section. There will be 2 tesing that will be reviewing and mention above which is the used of static camera and moving camera (support with Servo + Pan Tilt Control)
- Arduino File: Use this to control servo based on certain situation (moving camera or static camera)
- Face_detection_test.py: Testing the face tracking using haarcascade technique.
- Smart Air Cooler.py: Final code that need to connect with arduino module in order to work (Noted: It is important to select correct .ino file to avoid any damage to servo)
On the next section.
on this section we will explain on how we build this project and how the code are implemented. we will divided the part and split it into certain categories. The first part on how to make a cooler, the second is the electornic device (Servomachanism) and finally is the software used for coding in order to create an AI that control the device.
Cooler Generator Part
The component used to build this part are :
- 2x 150mm Fan
- 1x 92mm Fan (with heatsink)
- 1x Radiator
- 1x Waterpump (with reservoir)
- 1x powersupply (12v 10A)
Servomachanism
The component used to build this part are :
- 2x Servo MG90s
- 1x Servo Pan Tilt Platform
- 1x Arduino Mega
- 1x USB Webcam
- 1x breadboard (used for connection both Servomachanism and Cooler)
Software
IDE used:
- Arduino IDE
- Pyhton (AI)
For the final touch of this project, we will combine all part together becoming one full build project. but to take noted also that even it is the final result (prototype), due to pandemic of Covid-19 happen while this project are ongoing. Some assembly item may be missing like the servo should be equip with the fan like the design above.
This problem is occur due to money issue and the limitation to do a reaseach on the outside of the campus to find the part to build for the fan blower and also for the container. Because of that we only used available item only to demonstrate on the project will works. As a student and also due to virus pendemic, we have a hard time to communicate with team especially this project is IoT project based that require a lot of team face-to-face meeting. This what we have to deliver or present of the final project and supposedly we should make a box and putting all the component inside the container.
so special thank to our lecture for giving this opportunity to learn Artificial Intelligent, and same goes to my team for the support. For more information we will make a video presentation on the section below.
In this video, we will be presenting our final AI project which is a Smart Air Cooler using Python, OpenCV, haarcascades and combine it with arduino to make it an IoT project. To create our Smart Air Cooler, we use haarcascades which is a simple dataset that have already trained of thousand of people face. We fine-tuned using arduino code to move the servo on that being equip with any camera or webcam available
It is still in prototype and we hope to able to finish this final product.
Smart Air Cooler (Click image bottom to watch video)
