A System for Determining Indoor Air Quality from Images of an Air Sensor Captured on Cell Phones

We have designed and implemented a system that analyzes images of air sensors taken on cell phones and extracts indoor air pollution information by comparing the sensor to a calibrated color chart. This system will be deployed with Project Surya, a pilot project of the UN Environmental Program that aims to replace traditional cooking methods with clean cooking technology in several rural Indian villages. In order to evaluate the effectiveness of the cooking technology at reducing indoor pollution, several hundred homes will receive a special air sensor that changes brightness based on the concentration of air pollution. Our system is capable of automatically getting data from multiple sensors each day by analyzing photographs of them. In order to operate in the presence of noise and inconsistent lighting conditions, our system employs a median sensor, finds the light intensity gradient across images, and calibrates the images if possible. We used a GVF active contour segmentation algorithm that can segment the two objects in the images (the sensor and the color chart) followed by a mode brightness matching algorithm to match the sensor and color chart. We tested our system by comparing results of manual analysis of seven different sensors to results obtained using our system. In these tests, our system was shown to match the sensor to a color on the chart with a maximum discrepancy of 1 bar on the color chart from the average result of manual analysis, with an average deviance of .406 of a bar.