Search Results (3)

It is a challenge to identify food waste sources in all-weather industrial environments, as variable lighting conditions can compromise the effectiveness of visual recognition models. This study proposes and validates a robust, interpretable, and adaptive multimodal logic fusion method in which sensor dominance is dynamically assigned based on real-time illuminance intensity. The method comprises two foundational components: (1) a lightweight MobileNetV3 + EMA model for image recognition; and (2) an audio model employing Fast Fourier Transform (FFT) for feature extraction and Support Vector Machine (SVM) for classification. The key contribution of this system lies in its environment-aware conditional logic. The image model MobileNetV3 + EMA achieves an accuracy of 99.46% within the optimal brightness range (120–240 cd m⁻²), significantly outperforming the audio model. However, its performance degrades significantly outside the optimal range, while the audio model maintains an illumination-independent accuracy of 0.80, a recall of 0.78, and an F1 score of 0.80. When light intensity falls below the threshold of 84 cd m⁻², the audio recognition results take precedence. This strategy ensures robust classification accuracy under variable environmental conditions, preventing model failure. Validated on an independent test set, the fusion method achieves an overall accuracy of 90.25%, providing an interpretable and resilient solution for real-world industrial deployment. Full article

To ensure the effective implementation of food waste reduction in college cafeterias, Capital Normal University developed an automatic plate recognition system based on machine vision technology. The system operates by obtaining images of plates (whether clean or not) and the diners’ faces through multi-directional monitoring, then employs several deep learning models for the automatic localization and identification of the plates. Face recognition technology links the identification results of the plates to the diners. Additionally, the system incorporates innovative educational mechanisms such as online feedback and point redemption to encourage student participation and foster thrifty habits. These initiatives also provide more accurate training samples, enhancing the system’s precision and stability. Our findings indicate that machine vision technology is suitable for rapid identification and location of clean plates. Even without optimized network parameters, the U-Net network demonstrates high recognition accuracy (MIOU of 68.64% and MPA of 78.21%) and ideal convergence speed. Pilot data showed a 13% reduction in overall waste in the cafeteria and over 75% user acceptance of the mechanism. The implementation of this system has significantly improved the efficiency and accuracy of plate recognition, offering an effective solution for food waste prevention in college canteens. Full article

Waste recycling is critical for the development of smart cities. Local authorities are responsible for the disposal of waste plastics, but the extent of material recycling is insufficient, and much of the waste generated is incinerated. This conflicts with the trend of decarbonisation. Of particular note are the effects of the COVID-19 pandemic, during and after which large quantities of waste plastics, such as plastic containers and packaging, were generated. In order to develop a sustainable smart city, we need an effective scheme where we can separate materials before they are taken to the local authorities and recyclers. In other words, if material identification can be performed at the place of disposal, the burden on recyclers can be reduced, and a smart city can be created. In this study, we developed and demonstrated smart material identification equipment for waste plastic materials made of PET, PS, PP, and PE using GaP THz and sub-THz wavelengths. As basic information, we used a GaP terahertz spectrometer to sweep frequencies from 0.5 THz to 7 THz and measure the spectrum, and the transmittance rate was measured using the sub-THz device. The sub-THz device used a specific frequency below 0.14 THz. This is a smaller, more carriable, and less expensive semiconductor electronic device than the GaP. Moreover, the sub-terahertz device used in the development of this equipment is compact, harmless to the human body, and can be used in public environments. As a result, smart equipment was developed and tested in places such as supermarkets, office entrances, and canteens. The identification of materials can facilitate material recycling. In this study, we found that measuring devices designed to identify the PET and PS components of transparent containers and packaging plastics, and the PP and PE components of PET bottle caps, could effectively identify molecular weights, demonstrating new possibilities for waste management and recycling systems in smart cities. With the ability to collect and analyse data, these devices can be powerful tools for pre-sorting. Full article