2nd ICAI 2022

International Conference on Automotive Industry 2022

Mladá Boleslav, Czech Republic

elements of Industry 4.0 are Big Data and Analytics, Autonomous Robots, Simulation, Horizontal and Vertical System Integration, The Industrial Internet of Things, Cloud Computing, Additive Manufacturing, Augmented Reality and Cyber Security (Erboz, 2017). The described assumptions and Industry 4.0 elements are based on the acquisition and subsequent processing of large amounts of data, which should be of the highest possible quality. The collection of such data for logistics processes in the process improvement phases is time-consuming and costly. The fourth industrial revolution can be difficult to implement for small and medium-sized companies due to the costs for data acquisition and post-processing. In this article, we discuss a deep neural network with transfer learning approach that reduces implementation costs to a minimum. Observation of the functioning intralogistics system in order to improve it requires long term actions involving many resources. Using classical methods of process evaluation, it is difficult to obtain satisfactory results of current process efficiency monitoring. Activities requiring the involvement of engineering personnel in the ongoing analysis of the process in modern business conditions, are necessary to ensure the intellectual resources of the organization for the generation of competitive advantage. The following sections of the paper will present a case study of analysis of the possibility of improving the internal distribution process based on the classical logistic train using trained neural networks in the YOLO3 architecture fed with CCTV images. 2.1 Problem Formulation The case in paper concerns the detection of 3 classes of objects identified on the basis of camera images with a resolution of 1920 by 1080 pixels. The first 2 objects shown in Figure 1 represent a logistics train and a platform, while the 3rd one is a forklift carrying battery electrodes without the required protection. The research problem was the correct and real-time detection of the described objects. With correct detection, it is possible to develop solutions to continuously analyse the efficiency of a transport system based on non-autonomous means of transport in internal logistics, without the need to invest in additional infrastructure to collect data on the location and occupancy of means of transport. In addition, a solution not integrated with other IT systems allows to bypass unfavourable legal regulations requiring the payment of fixed integration fees. 2. Problem Formulation and Methodology

Figure 1: Case study class object examples

Source: Own elaboration

109