Development of an artificial vision system for the detection of free spaces in private parking lots in Lima (#390)

Authors

Orellana-Huamanchaqui, Kenedy

Abstract

Private parking lots face significant challenges due to the increasing number of vehicles, requiring advanced technological solutions to optimize space use and improve user experience. The artificial vision system for spot detection is proposed as an innovative solution using machine learning techniques. The VDI 2206 methodology was adopted due to its suitability in the design of mechatronic systems, integrating various areas of knowledge for effective system control. The development of the system was divided into five key stages. In the first stage, "Requirements", quality images were collected to define architectures and select electronic and software components. In the second stage, the hyperparameters of the VGG16 model and the SVM classifier were adjusted, designing flowcharts and programming. The third stage included coding the detection model, developing live video streaming modules, and improving image post-processing. The fourth stage focused on the assembly and configuration of the hardware and camera, optimizing the code for accurate real-time detection. Finally, in the fifth stage, the data were validated, and the model was evaluated, reaching a precision of 0.90 and an accuracy of 0.92 in image processing in one second, ensuring the coherence and effectiveness of the process and achieving effective detection of parking spaces.