This paper aims to develop a background subtraction algorithm based on Gaussian Mixture Model (GMM) using Probability Density Function (PDF) to identify the location of moving objects over a belt conveyor for pick and place operations using an industrial robot. In the present work, a stationary webcam is placed above the conveyor system to capture images of the objects that are coming into the view field. The objects of interest are identified by subtracting the background image (reference frame) from the current image frame based on the probability density function of respective pixels over time. The subtracted image frame is processed to extract the attributes such as location, colour, and shape of the objects. The extracted information, in turn, helps the robot to pick the desired object of interest. The results indicated that the GMM based background subtraction is more precisely extracting the features of the object than the direct subtraction technique for robotic applications. The algorithm is developed using MATLAB software.

本文旨在开发一种基于高斯混合模型（GMM）的背景减除算法，该算法使用概率密度函数（PDF）来识别用于工业机器人拾取和放置操作的皮带输送机上移动物体的位置。在当前的工作中，固定的摄像头被放置在传送带系统上方，以捕获进入视野的物体的图像。通过基于各个像素随时间的概率密度函数从当前图像帧中减去背景图像（参考帧）来识别感兴趣的对象。对减去的图像帧进行处理以提取属性，例如对象的位置，颜色和形状。提取的信息又帮助机器人选择所需的感兴趣对象。结果表明，与机器人应用程序中的直接减法技术相比，基于GMM的背景减法更精确地提取了对象的特征。该算法是使用MATLAB软件开发的。