A Block Recognition System Constructed by Using a Novel Projection Algorithm and Convolution Neural Networks

In the age of Industry 4.0, the techniques of artificial intelligence and pattern recognition play a critical role to develop the smart factories. In this paper, a block recognition system, named e-Block, was developed by using a novel projection algorithm and the convolution neural networks. The developed system displays a picture of the target object (e.g., a car or a house), and the children follow the instructions provided by the system and use the various blocks to build the object. Subsequently, this system compares the assembled block with the target object and determines whether the shape is identical. To identify the assembled block, this system applies Kinect to obtain information on the depth of the object and a new projection algorithm is proposed for converting the depth information into three feature images. By integrating three feature images, the convolution neural networks (CNN) are employed to construct the classifier to identify the assembled block. In the experiments conducted in this paper, the CNN classifier was compared with three classification algorithms. The experimental results show that the CNN classifier can accurately recognizes whether the assembled object is identical to the target object and outperform the compared classification algorithms. In additions, the experimental results also reveal that the proposed recognition algorithm can be a useful technique for applying in various applications of Industry 4.0.