An Efficient Classification Process using Supervised Deep Learning and Robot Positioning based on Embedded PD-FLC

Abstract

Recently, computer vision has grown considerably in roboticsarms such as packaging and sorting for Industrial applications. The real challenge is how to improve the present sorting system based on artificial intelligence (AI) in the four points to identify, manipulate, select, and sort objects depending on their features. Therefore, this paper proposes an automatic sorting system based on Deep Learning (DL) and control in a robotic arm using an embedded PD-FLC. The proposed algorithm is divided into three stages; The first stage introduces building a Model for the identification and classification of fruits using the concept of Supervised deep learning (SDL) using the convolution neural network (CNN) algorithm. The model is trained using our data set of images that consider 12 classes of fruits named (Fruits-dataset). The second stage uses the pre-trained model and a web camera to automatically identify and classify the detected objects into 12 categories of fruits in real-time. The third stage uses an embedded proportional derivative fuzzy logic controller (PDFLC) to control the position of a 3DOF robotic arm to locate the classified object in the desired location. Our proposed SDLCNN model is tested to classify the fruits under different environment states (outdoor and indoor) in different color modes and intensities. The proposed SDL-CNN algorithm is compared with different state-of-the-art methods. Correspondingly, the obtained results show the effectiveness, high accuracy, and low cost of the developed design with the capability of real-time identifying and classifying the fruits.