Overview / Usage

Development of Smart Traffic Lights (STL) for both efficient traffic management and pedestrians safety and convenience is most critical in realization of any smart city project. The timing of these STL is controlled dynamically on the basis of vehicular and pedestrian traffic. Many implementations of STL that makes use of color cameras and other sensors have been proposed. However, these are in the early stages of development and have severe performance issues under poor lighting and weather conditions. On the other hand, thermal imaging have been successfully used in surveillance applications and have given promising results both under poor lighting and weather conditions. With the decreasing cost of these images, their applications in different domains such as pedestrian detection have become possible. Thermal imaging has been chosen due to the following advantages as it can be used in low light conditions unlike normal vision cameras, comparatively less installation cost as compared to pressure plates, works on fluctuating ranges of temperature and Immune to headlight and sunlight glare The aim of the project is to explore the various devices and techniques which can be used for efficient pedestrian detection from the captured image.

Methodology / Approach

In this project, the thermal camera is placed at traffic light, from the traffic light data is collected and send to the server. At the server, data is processed and frames are extracted from the video. From these extracted frame the pedestrians are detected and a count is generated depending on the number of pedestrians detected. This count is compared with a set threshold, if count is greater than threshold green LED will glow and if less than threshold red LED will glow.

Technologies Used

OpenCV
CNN
Keras
Python
Tensorflow