Contact us

Откроется новое окно Откроется новое окно

Rapid pedestrian detection: Advanced algorithms and hardware accelerators for ADAS

Rapid pedestrian detection: Advanced algorithms and hardware accelerators for ADAS

CoHOG is a proprietary algorithm that was developed by Toshiba for the purpose of detecting objects that concern an ADAS system - and a driver. Its effectiveness comes from examining the edges associated with humans including elbows, shoulders, legs and hips.

The gradients and positions of these edges are dictated by human physiological parameters and must be situated within the confines of these limits for the object to be considered 'human'. Rather than a single gradient in isolation, CoHOG looks for pairs of oriented gradients allowing superior human detection to be achieved.

The process is highly processor-intensive and is ideally suited to the parallel processing abilities of a device such as Toshiba’s TMPV7502. This processor incorporates five hardware accelerators. Each accelerator has a specific role to play within the image-processing pipeline, all acting upon the input from the camera module in real-time.

Affine Accelerator: Scales images so Regions of Interest (ROIs) can be compared. It also removes distortion effects from wide-angle lenses.

Filter Accelerators: For filter operations such as edge detection, noise reduction or colour space conversion.

Histogram Accelerator: Calculates the histogram for camera images, allowing light intensity processing.

CoHOG Accelerator: Recognition of shapes and edges of obstacles.

Matching Accelerator: Tracks objects from one frame to the next, allowing ADAS to determine if the object is within, or likely to move within, the trajectory of the vehicle.

To learn how the TMPV7502 can accelerate your ADAS system design, please click here:

Click here to download the whitepaper

To Top
·Before creating and producing designs and using, customers must also refer to and comply with the latest versions of all relevant TOSHIBA information and the instructions for the application that Product will be used with or for.