Indian car safety system developed to prevent collisions with sacred cows
Scientists based at Gujarat Technological University, India, have developed a real-time car safety system which detects nearby cows and prompts the driver to brake if a collision is likely. This takes inspiration from the safety features in driverless cars to create a low-cost system that could be installed in standard vehicles.
India has the second-largest road network and the highest number of road accidents and fatalities in the world. While its people continue to become richer and acquire more cars, these incidents are likely to rise.
According to Sachin Sharma and Dharmesh Shah, researchers based at Gujarat Technological University, road maintenance in India is struggling to keep up with demand, particularly in rural areas where roads connecting towns and villages are often in need of repair. On poor roads such as these, driving can be very dangerous and collisions with people, animals and other vehicles is commonplace.
Taking inspiration from the extensive automatic safety features developed for driverless cars, Sharma and Shah were prompted to develop a cheap, simple safety system that would detect obstacles – specifically cows – and prevent accidents from occurring.
In Hindu-majority India, cows are respected and accorded freedom to roam. Slaughter of cattle is illegal in most states of India and recently the state of Gujurat extended the punishment for cow slaughter from seven years to life imprisonment.
The researchers’ cow-detection algorithm combined two object detection methods: a histogram of oriented gradients, often used to detect people in video footage, and boosted cascade classifiers, which are used in facial recognition software.
A simple dashboard camera collects live video of the road ahead and the algorithm searches for signs of cows in the frames. The algorithm detects whether an object is a cow and, if so, calculates the distance of the animal from the car based on its size and the vehicle’s speed. If necessary, the system then alerts the driver to decelerate or brake.
During their study, the researchers tested their new system on video footage featuring 80 cows. The algorithm achieved a cow detection rate of approximately 80 per cent, with a low false-detection rate.
Training and testing the system on larger data sets with different angles and weather conditions will further improve the detection rate, the researchers report in their paper, published in the International Journal of Vehicle Autonomous Systems. They suggest that the method could then be extended to detect cows at night and eventually to detect other animals.
Sign up to the E&T News e-mail to get great stories like this delivered direct to your inbox every day.