Child Left in the Car Detection: Image Enhancement for Day and Night

Despite the government's awareness campaign on the safety of children in non-moving vehicles, the cases of children being trapped and suffocated in unattended cars keep rising. Children are often being ignored by their parents with the engine off. Furthermore, there are also children being ignored due to unstoppable calls from work. Next, existing systems with only 1 detection used have a limit either detecting face or signal detection. Most of the software tools have a limitation in detecting human face in a low light situation. Other than that, the obstacle like hand will decrease the accuracy of face detection. This project aims to develop a complete and adequate face detection system for detecting the presence of children in the car by detecting human physical features. The objective is to enhance the visualization of images for human identification and to measure the performance of the features selections for detection system. There are 100 sample images of child faces being collected, and three filters are being compared for image enhancement: fastnimeandenoisingcolored, histogram equalizer and median filter. During normal daylight, fastnimeandenoisingcolored achieves the highest percentage of accuracy of face and hand detection with 90%, followed by without filter and Median Filter with 89% accuracy. During night, 100 to 150 value of dimmer images, Histogram Equalizer achieves the highest accuracy percentage of face and hand detection with 85%, followed by fastnimeandesnoisingcolored and without filter with 81% accuracy. In summary, the face detection technology was able to recognize the child who had been left in the car and so aid in reducing the accident that occurred today.