Object Detection Using Computer Vision Methods on Real-Time Lux Sensor Data

Computer vision methods are conventionally used for identification of object boundaries from an image of interest. Here, it has been performed extensive performance analysis of computer vision techniques for detection of objects present in a room from the acquired real-time lux sensor data. A server-client wireless network is initially set up to capture horizontal lux data of a room using unmanned surface vehicle. The client unit comprises of light intensity sensor and IR transceiver module for capturing the positional lux information of an indoor space. The sensing block is interfaced with a wireless enabled microcontroller and it transmits the sensor values to a sever unit with the help of a wireless router. Now, the obtained values are stored in a computer system working as server module. The acquired real-time positional lux data are converted into image and different computer vision techniques e.g. Canny method, Prewitt method, Sobel method, and Laplacian of Gaussian method have been used to detect the possible position and object shape in the indoor space. Comparative assessment based on performance indices like recall, precision, and F1 value has been carried out to find out the most acceptable solution.