RESUMEN

The utilization of computer vision in smart farming is becoming a trend in constructing an agricultural automation scheme. Deep learning (DL) is famous for the accurate approach to addressing the tasks in computer vision, such as object detection and image classification. The superiority of the deep learning model on the smart farming application, called Progressive Contextual Excitation Network (PCENet), has also been studied in our recent study to classify cocoa bean images. However, the assessment of the computational time on the PCENet model shows that the original model is only 0.101s or 9.9 FPS on the Jetson Nano as the edge platform. Therefore, this research demonstrates the compression technique to accelerate the PCENet model using pruning filters. From our experiment, we can accelerate the current model and achieve 16.7 FPS assessed in the Jetson Nano. Moreover, the accuracy of the compressed model can be maintained at 86.1%, while the original model is 86.8%. In addition, our approach is more accurate than ResNet18 as the state-of-the-art only reaches 82.7%. The assessment using the corn leaf disease dataset indicates that the compressed model can achieve an accuracy of 97.5%, while the accuracy of the original PCENet is 97.7%.

Asunto(s)

Agricultura , Compresión de Datos , Granjas , Fenómenos Físicos , Automatización

RESUMEN

In optical camera communications (OCC), the provision of both flicker-free illumination and high data rates are challenging issues, which can be addressed by utilizing the rolling-shutter (RS) property of the image sensors as the receiver (Rx). In this paper, we propose an RS-based multiple-input multiple-output OCC scheme for the Internet of things (IoT) application. A simplified design of multi-channel transmitter (Tx) using a 7.2 × 7.2 cm2 small 8 × 8 distributed light emitting diode (LED) array, based on grouping of LEDs, is proposed for flicker-free transmission. We carry out an experimental investigation of the indoor OCC system by employing a Raspberry Pi camera as the Rx, with RS capturing mode. Despite the small area of the display, flicker-free communication links within the range of 20-100 cm are established with data throughput of 960 to 120 bps sufficient for IoT. A method to extend link spans up to 1.8 m and the data throughput to 13.44 kbps using different configurations of multi-channel Tx is provided. The peak signal-to-noise ratio of ~14 and 16 dB and the rate of successfully received bits of 99.4 and 81% are measured for the shutter speeds of 200 and 800 µs for a link span of 1 m, respectively.

RESUMEN

In this Letter, we develop a novel technique, to the best of our knowledge, to increase the link span ($ {L_s} $Ls) of a rolling shutter (RS)-based optical camera communications (OCC) system by reducing the spatial bandwidth of the camera in the out-of-focus regions. We demonstrate a 400 m line-of-sight RS-based OCC link, which is to date the longest $ {L_s} $Ls reported in these systems, and develop a detection method to extract the information out of the video frames, successfully. The proposed system relaxes the condition of a large surface area for the transmitter light source. Consequently, we show that at 400 m $ {L_s} $Ls and exposure times of 100-80 µs, a data rate of 450 bps is achieved successfully.

RESUMEN

The importance of detection of parking space availability is still growing, particularly in major cities. This paper deals with the design of a distributed wireless camera system for the management of parking spaces, which can determine occupancy of the parking space based on the information from multiple cameras. The proposed system uses small camera modules based on Raspberry Pi Zero and computationally efficient algorithm for the occupancy detection based on the histogram of oriented gradients (HOG) feature descriptor and support vector machine (SVM) classifier. We have included information about the orientation of the vehicle as a supporting feature, which has enabled us to achieve better accuracy. The described solution can deliver occupancy information at the rate of 10 parking spaces per second with more than 90% accuracy in a wide range of conditions. Reliability of the implemented algorithm is evaluated with three different test sets which altogether contain over 700,000 samples of parking spaces.

RESUMEN

The automatic observation of the night sky through wide-angle video systems with the aim of detecting meteor and fireballs is currently among routine astronomical observations. The observation is usually done in multi-station or network mode, so it is possible to estimate the direction and the speed of the body flight. The high velocity of the meteorite flying through the atmosphere determines the important features of the camera systems, namely the high frame rate. Thanks to high frame rates, such imaging systems produce a large amount of data, of which only a small fragment has scientific potential. This paper focuses on methods for the real-time detection of fast moving objects in the video sequences recorded by intensified TV systems with frame rates of about 60 frames per second. The goal of our effort is to remove all unnecessary data during the daytime and make free hard-drive capacity for the next observation. The processing of data from the MAIA (Meteor Automatic Imager and Analyzer) system is demonstrated in the paper.

RESUMEN

Meteor detection and analysis is an essential topic in the field of astronomy. In this paper, a high-sensitivity and high-time-resolution imaging device for the detection of faint meteoric events is presented. The instrument is based on a fast CCD camera and an image intensifier. Two such instruments form a double-station observation network. The MAIA (Meteor Automatic Imager and Analyzer) system has been in continuous operation since 2013 and has successfully captured hundreds of meteors belonging to different meteor showers, as well as sporadic meteors. A data processing pipeline for the efficient processing and evaluation of the massive amount of video sequences is also introduced in this paper.