Assertive Vision

Overview

Apical’s Assertive Vision processor is a state-of-the-art detection, classification and tracking engine capable of accurate analysis of people and other objects in real time. It is designed for integration into embedded systems and requires no external software processing.

It makes no compromise on performance and processes all pixels in HD video in real time. For example, any number of people can be detected and tracked based on a set of models of the face, upper and full body.

It can run independently of Apical’s Assertive Camera image signal processor but takes advantage of Assertive Camera’s image processing functions to enhance accuracy when used in conjunction.

The processor is implemented on the Xilinx Zynq 7045 SoC and is designed for integration into FPGA and SoC devices.

Assertive Vision

Key Features

System comprises a dedicated hardware analytics engine together with real-time firmware running on an embedded processor.

  • Detects and classifies objects as small as 45 pixels wide in a full HD frame
  • Real-time processing of full HD video without resize
  • Up to 16 different object types simultaneously detected and classified
  • Any number of object detections per frame
  • Real-time grouping and tracking of any number of individual objects
  • Works directly on raw sensor data or on standard camera output
  • Currently trained to detect and track people in complex environments
  • Full hardware implementation of main detection and classification algorithms

The system is currently implemented on the Xilinx Zynq 7045 SoC development board.

street_1