Correction of Non-ideal Lighting Conditions for Face Recognition in Alcohol Ignition Interlock Devices

The alcohol interlock system manufactured by Alcohol Countermeasure Systems (ACS) is an in-vehicle breath alcohol testing instrument that prevents a vehicle engine from being started if the breath alcohol concentration of the driver is over a pre-set limit. With the current countermeasures, the alcohol interlock devices, with a high certainty, can confirm that the blown air is from a human, but fails to indicate whether he/she is the intended driver. A fundamental problem exists in the current alcohol interlocks devices is that they lack a “smart” monitoring system to know the identity of the person who is using the device. Implementing a facial recognition system within these devices is the key; however, experiments show that images of ‘poor quality’ such as driver’s pose variation, image blurring due to vehicle bumping and vibration, and illumination variations negatively and dramatically affected the performance of facial recognition. Therefore, the integration of illumination correction methods into ACS facial recognition system becomes the key to improving the performance of the alcohol interlock system to a satisfactory level. Specifically, the objectives of the proposed research are: a) to investigate facial recognition under non- ideal conditions such as non-uniform illumination, shadow artifacts and glare vision; b) to develop a new prototype interlock device with capabilities to accurately, automatically, and continuously authenticate the car driver; c) to transfer new knowledge acquired and new technology developed in the project to ACS, and other Canadian industry specialized in public safety, and intelligent systems; and d) to train highly qualified personnel for Canada’s R&D industry in the public safety sector and intelligent systems.

Faculty Supervisor:

Drs. Ling Guan, A.N. Venetsanopoulos

Student:

Naimul Mefraz Khan, Wael Louis, Xiaoming Nan, Fei Guo

Partner:

Alcohol Countermeasure Systems Corp.

Discipline:

Engineering - computer / electrical

Sector:

Manufacturing

University:

Ryerson University