Advanced SURF Features Based Flexible Object Detection

Algorithm Structure

Abstract

In the background of industrial automation trend becoming more and more common, Flexible products produced on modern assembly line. like cloths or rubber, are available to each other or stack on each other. The product efficiency will be severely influenced. Generally speaking, the current methods are on single feature like color or shape works well on non - flexible product; Or we can count and identify positions and postures of multiple work piece by obtaining 3D point cloud information of the work piece, but point cloud information cannot exclude debris in the field of vision, if we need to keep flexible products stable, then extra flatten process is needed, which bring more costs. Considering the problems in the actual industrial production environment, this study plans to solve the following bottleneck problems in related fields by studying the detection algorithm based on the monocular machine vision with the lowest relative cost. The detection algorithm based on the monocular machine vision with the lowest relative cost. 1. The problem of object recognition of flexible parts with distorted posture and mutual folding under different lighting conditions; 2. The problem of precise counting and position ordering of cluttered stack artifacts based on vision; 3. When different types of flexible parts are stacked together, the problem of debris elimination and the problem of accurate counting and position and posture ordering of target work piece.

Publication
Robio 2019

Supplementary notes can be added here, including code, math, and images.

Yunfu Deng
Yunfu Deng
Research Assistant

My research interests