A new pedestrian detection system has been created by electrical engineers from the University of California in San Diego (UCSD) in the US.

The real-time technology is capable of detecting pedestrians between two and four frames every second and with better accuracy compared with existing systems.

The detection system can be installed in robotics, ‘smart’ vehicles, and image and video search systems.

"We’re aiming to build computer vision systems that will help computers better understand the world around them."

The latest pedestrian detection algorithm is developed by a team led by electrical engineering professor Nuno Vasconcelosa of UCSD’s Jacobs School of Engineering.

Professor Vasconcelos and his team have combined cascade detection, which is conventional computer vision classification architecture, with deep learning models.

Vasconcelos said: "We’re aiming to build computer vision systems that will help computers better understand the world around them."

The pedestrian detection system can break down an image into small windows, which are then processed by a classifier to detect the presence or absence of a pedestrian.

The system generates images of pedestrians in different sizes, based on their distance to the camera and locations within an image.

In cascade detection system, the detector works in a number of stages, which adds to its complexity.

The team thus developed a new algorithm that can address these issues.

The new algorithm includes deep learning models in the final stages of a cascaded detector, which process only fewer windows and are well-equipped to make complex pattern recognition.

Even though they work well for the final stages of the cascade detection system, the deep learning models are sufficiently complex to be used in the early cascade stages.

Therefore, the team developed a new cascade architecture that combines classifiers from different families, including simple classifiers from the early stages and deep learning models from the later ones.

Vasconcelos added: "No previous algorithms have been capable of optimising the trade-off between detection accuracy and speed for cascades with stages of such different complexities.

"In fact, these are the first cascades to include stages of deep learning. The results we’re obtaining with this new algorithm are substantially better for real-time, accurate pedestrian detection."