This work copes with the design and implementation of a wireless sensors network architecture to automatically and continuously monitor, for the first time, the manufacturing process of Sardinian Carasau bread. The case of a traditional bakery company facing the challenge of the Food-Industry 4.0 competitiveness is investigated. The process was analyzed to identify the most relevant variables to be monitored during the product manufacturing. Then, a heterogeneous, multi-tier wireless sensors network was designed and realized to allow the real-time control and the data collection during the critical steps of dough production, sheeting, cutting and leavening. Commercial on-the-shelf and cost-effective integrated electronics were employed, making the proposed approach of interest for many practical cases. Finally, a user-friendly interface was provided to enhance the understanding, control and to favor the process monitoring. With the wireless senors network (WSN) we designed, it is possible to monitor environmental parameters (temperature, relative humidity, gas concentrations); cinematic quantities of the belts; and, through a dedicated image processing system, the morphological characteristics of the bread before the baking. The functioning of the WSN was demonstrated and a statistical analysis was performed on the variables monitored during different seasons.

In this paper, we address the problem of automatic image segmentation methods applied to the partial automation of the production process of a traditional Sardinian flatbread called pane Carasau for assuring quality control. The study focuses on one of the most critical activities for obtaining an efficient degree of automation: the estimation of the size and shape of the bread sheets during the production phase, to study the shape variations undergone by the sheet depending on some environmental and production variables. The knowledge can thus be used to create a system capable of predicting the quality of the shape of the dough produced and empower the production process. We implemented an image acquisition system and created an efficient machine learning algorithm, based on support vector machines, for the segmentation and estimation of image measurements for Carasau bread. Experiments demonstrated that the method can successfully achieve accurate segmentation of bread sheets images, ensuring that the dimensions extracted are representative of the sheets coming from the production process. The algorithm proved to be fast and accurate in estimating the size of the bread sheets in various scenarios that occurred over a year of acquisitions. The maximum error committed by the algorithm is equal to the 2.2% of the pixel size in the worst scenario and to 1.2% elsewhere.

This work deals with the study and design of a wireless sensors network (WSN) architecture to optimize the manufacturing process of Sardinian Carasau bread. In a scenario of a small-scale bakery industry, a set of process variable to be monitored to optimize the product manufacturing is defined. Then a WSN architecture is studied and set-up to allow for real-time and in situ data collection during both the dough production and the leavening steps. Commercial on-the-shelf and cost-effective integrated electronics is employed. Finally, a user-friendly interface is provided to enhance the understanding, control and, in particular, to favor the optimization of the process.

The ubiquitous nature and great potential of Wireless Sensors Network has not yet been fully exploited in automotive applications. This work deals with the choice of the cost-effective hardware required to face the challenges and issues proposed by the new trend in the development of intelligent transportation systems. With this aim, a preliminary WSN architecture is proposed. Several commercially available open-source platforms are compared and the Raspberry Pi stood out as a suitable and viable solution. The sensing layer is designed with two goals. Firstly, accelerometric, temperature, and relative humidity sensors were integrated on a dedicated PCB to test if mechanical or environmental stresses during bus rides could be harmful to the device or to its performances. The physical quantities are monitored automatically to alert the driver, thus improving the quality of service. Then, the rationale and functioning of the management and service layer is presented. The proposed cost-effective WSN node was employed and tested to transmit messages and videos, while investigating if any quantitative relationship exists between these operations and the environmental and operative conditions experienced by the hardware.

Interpretable machine learning for ordinal quality grading faces a fundamental tension between model transparency and predictive performance, particularly under small-data conditions where end-to-end deep learning is unreliable and domain knowledge must compensate for limited training samples. We present a dual-target feature engineering framework for interpretable ordinal grading validated on pane Carasau, a traditional flatbread whose extreme surface variability makes it a challenging small-data benchmark for machine learning under realistic acquisition constraints. The pipeline extracts 116 handcrafted visual descriptors organised into four families—colour, texture, spatial, and hotspot—and grades the quality along two independent axes: global toasting intensity and spatial uniformity, complemented by a continuous Toasting Index for process monitoring, on a dataset of 1512 images spanning four acquisition campaigns and three product types. On the primary within-batch evaluation set Campaign 01, $N=1090$), XGBoost achieves F1 macro $=0.906$ for toast classification and $R^2=0.886$ for continuous regression, substantially outperforming two fine-tuned CNN baselines on the same evaluation set (MobileNetV2: F1 $=0.523$; EfficientNet-B0: F1 $=0.518$). Feature importance analysis reveals that colour descriptors dominate toasting prediction ($87.5\%$), whilst spatial and texture features are essential for uniformity assessment ($47.4\%$ combined), providing physically grounded explanations directly traceable to the underlying thermal process. Cross-batch generalisation on held-out campaigns is moderate for the same product (XGB F1 = 0.718, $\kappa$ = 0.703); cross-product transfer to geometrically distinct variants requires product-specific adaptation. The framework requires no GPU, runs on standard CPU hardware at 4 s per image, and provides complete decision transparency, supporting deployment without specialised hardware.

Automated visual quality inspection of artisanal baked goods requires methods that are accurate, interpretable, and capable of distinguishing multiple independent quality dimensions from a single image. We present a dual-target feature engineering framework for ordinal grading of pane Carasau, a traditional Sardinian flatbread with high surface variability, validated on a dataset of 1,512 images spanning four acquisition campaigns and three product types. The pipeline extracts 116 handcrafted visual descriptors organised into four families—colour, texture, spatial, and hotspot—and gradesquality along two independent axes: global toasting intensity and spatialuniformity, complemented by a continuous Toasting Index for process monitoring. On the primary within-campaign set (N = 1,090), XGBoost achievesF1 macro = 0.906 for toast classification and R2 = 0.886 for continuousregression, substantially outperforming two fine-tuned CNN baselines (MobileNetV2: F1 = 0.524; EfficientNet-B0: F1 = 0.518). Feature importance analysis confirms that colour descriptors dominate toasting prediction (87.5%), whilst spatial and texture features are essential for uniformity assessment (46.0% combined). Cross-campaign evaluation on three held-out sets reveals moderate cross-batch generalisation on the same product (XGB F1 = 0.718, κ = 0.703), with colour and texture features proving more transferable than spatial descriptors; cross-product transfer to geometrically distinct bread requires product-specific adaptation. The framework requiresno GPU, runs on standard CPU hardware at 4 s per image, and provides complete decision transparency traceable to physical process parameters.