The current roadway monitoring is expensive and not systematic. This paper proposes a new system able to evaluate the pavement quality of road infrastructure. The embedded system records and processes the acoustic data of the wheel-road interaction and classifies in real-time roadways’ health thanks to integrated AI solutions. The measurements to produce the dataset to train a convolutional neural network (CNN) were collected using a vehicle operating at different cruise speeds in the area of Pisa. The dataset is composed by acoustic data belonging to several typologies of roads: dirty or grass roads, high roughness surfaces and roads with cracks or potholes. The raw audio signals were split, labelled, and converted into images by calculating the Mel spectrogram. Finally, the authors designed a tiny CNN with a size equal to 18 kB able to classify between four different classes: good quality road, ruined road, silence and unknown. The CNN architecture achieves an accuracy of about 93% on the original model and 90% on the quantized one. Quantization permits to convert the final architecture into a suitable form to be deployed on a low-complex embedded system integrated in the tyre cavity. In addition, a custom board was designed to act as IoT node thanks to a Bluetooth Low Energy communication towards smartphones and/or car infotainment systems. These systems, featured with GPS, guarantee to obtain real-time maps service of road quality. At authors’ knowledge, this is the first real-time and fully integrated solution at the state of the art for road pavement quality analysis and classification on acoustic data.
19 March 2021 in Journal of Real-Time Image Processing
Smoke detection represents a critical task for avoiding large scale fire disaster in industrial environment and cities. Including intelligent video-based techniques in existing camera infrastructure enables faster response time if compared to traditional analog smoke detectors. In this work presents a hybrid approach to assess the rapid and precise identification of smoke in a video sequence. The algorithm combines a traditional feature detector based on Kalman filtering and motion detection, and a lightweight shallow convolutional neural network. This technique allows the automatic selection of specific regions of interest within the image by the generation of bounding boxes for gray colored moving objects. In the final step the convolutional neural network verifies the actual presence of smoke in the proposed regions of interest. The algorithm provides also an alarm generator that can trigger an alarm signal if the smoke is persistent in a time window of 3 s. The proposed technique has been compared to the state of the art methods available in literature by using several videos of public and non-public dataset showing an improvement in the metrics. Finally, we developed a portable solution for embedded systems and evaluated its performance for the Raspberry Pi 3 and the Nvidia Jetson Nano.
This paper proposes a video-based smoke detection technique for early warning in antifire surveillance systems. The algorithm is developed to detect the smoke behavior in a restricted video surveillance environment, both indoor (e.g., railway carriage, bus wagon, industrial plant, or home/office) or outdoor (e.g., storage area or parking area). The proposed technique exploits a Kalman estimator, color analysis, image segmentation, blob labeling, geometrical features analysis, and M of N decisor, in order to extract an alarm signal within a strict real-time deadline. This new technique requires just a few seconds to detect fire smoke, and it is 15 times faster compared to the requirements of fire-alarm standards for industrial or transport systems, e.g., the EN50155 standard for onboard train fire-alarm systems. Indeed, the EN50155 considers a response time of at least 60 s for onboard systems. The proposed technique has been tested and compared with state-of-art systems using the open access Firesense dataset developed as an output of a European FP7 project, including several fire/smoke indoor and outdoor scenes. There is an improvement of all the detection metrics (recall, accuracy, F1 score, precision, etc.) when comparing Advanced Video SmokE Detection (AdViSED) with other video-based antifire works recently proposed in literature. The proposed technique is flexible in terms of input camera type and frame size and rate and has been implemented on a low-cost embedded platform to develop a distributed antifire system accessible via web browser.
This paper presents the development of a hardware/software system for the characterization of the electronic response of optical (camera) sensors such as matrix and linear color and monochrome Charge Coupled Device (CCD) or Complementary Metal Oxide Semiconductor (CMOS). The electronic response of a sensor is required for inspection purposes. It also allows the design and calibration of the integrating device to achieve the desired performance. The proposed instrument equipment fulfills the most recent European Machine Vision Association (EMVA) 1288 standard ver. 3.1: the spatial non uniformity of the illumination ΔE must be under 3%, and the sensor must achieve an f-number of 8.0 concerning the light source. The following main innovations have achieved this: an Ulbricht sphere providing a uniform light distribution (irradiation) of 99.54%; an innovative illuminator with proper positioning of color Light Emitting Diodes (LEDs) and control electronics; and a flexible C# program to analyze the sensor parameters, namely Quantum Efficiency, Overall System Gain, Temporal Dark Noise, Dark Signal Non Uniformity (DSNU1288), Photo Response Non-Uniformity (PRNU1288), Maximum achievable Signal to Noise Ratio (SNRmax), Absolute sensitivity threshold, Saturation Capacity, Dynamic Range, and Dark Current. This new instrument has allowed a camera manufacturer to design, integrate, and inspect numerous devices and camera models (Necta, Celera, and Aria).