Computer Vision Based Industrial Robotic Arm for Sorting Objects by Color and Height

Abstract

In industrial production systems, manufacturers often face difficulties in sorting different types of objects. Color and height-based sorting which is done manually by human is quite a tedious task and its needs countless time as well. For manual sorting, many workers are required, which can be quite expensive. Moreover, robots that can sort only color or height can’t be effective when there is a need of products with same color with different heights and vice versa. In this paper, a computer vision based robotic sorter is proposed, which is capable of detecting and sorting objects by their colors and heights at the same time. This work isn’t done before as height sorting of same shapes is a new technique, which is done with color sorting techniques by computer vision. It is equipped with a robotic arm having 6 degree of freedom (DOF), by which it picks up and then place objects according to its color and height, to a predetermined place as per the production system requirement. A camera with the computer vision software detects various colors and heights. Haar Cascade algorithm has been used to sort the products. This multi-DOF robotic sorter can be a remarkably useful tool for automating the production process completely, where multiple conveyor belts are used, which can reduce time complexity as well. In the proposed system, the efficiency of color and height sorting is around 99%, which proves the efficiency of our system. The overall improvement in the efficiency of the production process can be significantly enhanced by using this system.

Optimizing Robotic Arm Precision With Real-Time Machine Vision Feedback

“…The description of the arm manipulator, the subjects under consideration, the location of the image-capturing element for the implementation of computer vision, the type of image processing, the integration of the processing, and the movement due to it are all the main factors covered. Image processing for the arm's working [1,4,5,10] is the first step of the implementation. The segmentation of an image to understand its environment and feature at the pixel level is the core idea of processing the Image.…”

Real-Time Object Detection and Grasping for a Robotic Arm

“…Several related works use object detection and a robotic arm to perform a task. This work includes [4], [7], [10], & [11], where the robotic arm attempts to grasp objects when detected.…”

“…Using a deep neural network(DNN) 's capability to learn, observe, and make a judgment with a robotic arm's ability to grasp and physically perform a task is an attractive field to be explored. Many related works such as [4], [7], [10], & [11] have done such, deploying a robotic arm to perform an objective. [7] has a robotic arm that uses computer vision to sort objects.…”

“…Many related works such as [4], [7], [10], & [11] have done such, deploying a robotic arm to perform an objective. [7] has a robotic arm that uses computer vision to sort objects. A camera at the base lets the robot sort objects based on color and height when presented in front of the camera.…”

“…A fixed positioned camera compensates for the lack of information such as distance. By having a fixed camera in [7], the system will assume objects placed on the plane viewed by overhead, the worktable.…”