• IEMEK Journal of Embedded Systems and Applications
• /
• v.19 no.2
• /
• pp.73-81
• /
• 2024

Recently, mobile manipulators have been utilized in agriculture industry for weed removal and harvest automation. This paper proposes a domain adaptive fruit detection method for harvest automation, by utilizing OWL-ViT model which is an open-vocabulary object detection model. The vision-language model can detect objects based on text prompt, and therefore, it can be extended to detect objects of undefined categories. In the development of deep learning models for real-world problems, constructing a large-scale labeled dataset is a time-consuming task and heavily relies on human effort. To reduce the labor-intensive workload, we utilized a large-scale public dataset as a source domain data and employed a domain adaptation method. Adversarial learning was conducted between a domain discriminator and feature extractor to reduce the gap between the distribution of feature vectors from the source domain and our target domain data. We collected a target domain dataset in a real-like environment and conducted experiments to demonstrate the effectiveness of the proposed method. In experiments, the domain adaptation method improved the AP50 metric from 38.88% to 78.59% for detecting objects within the range of 2m, and we achieved 81.7% of manipulation success rate.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.34 no.6C
• /
• pp.642-648
• /
• 2009

In this paper, we use SA (Seysen's Algorithm) instead of LLL (Lenstra-Lenstra-Lovasz) to perform LRA (Lattice Reduction-Aided) detection. By using SA, the complexity of lattice reduction is reduced and the detection performance is improved Although the performance is improved using SA, there still exists a gap in the performance between SA-LRA and ML detection. To reduce the performance difference, we apply list of candidates scheme to SA-LRA. The list of candidates scheme finds a list of candidates. Then, the candidate with the smallest squared Euclidean distance is considered as the estimate of the transmitted signal. Simulation results show that the SA-LRA detection learn to quasi-ML performance. Moreover, the efficiency of the SA is shown to highly improve the channel matrix conditionality.

• Korean Journal of Environmental Agriculture
• /
• v.30 no.4
• /
• pp.440-445
• /
• 2011

BACKGROUND: This study was conducted to monitor the current status of pesticide residues and to assess their risk in domestic agricultural products. The samples were rice, barley, lettuce and perilla leaf. These four types of agricultural products were those with GAP(Good Agricultural Practice) certification, organic agricultural products, pesticide-free agricultural products or general agricultural products. METHODS AND RESULTS: They were purchased from traditional markets and supermarkets of 12 regions in Korea from July to August 2010. The total number of samples was 259 for agricultural products and these were analyzed by GC/ECD, GC/NPD and GC/MSD. We used multiresidue methods to analyze for 110 different pesticides except for herbicides. CONCLUSION: In this study, residual pesticides were detected in 18 samples. Among these general agricultural products, organic agricultural products and products with GAP-certification were detected in 12, 4 and 2 samples, respectively. Detection rates of general agricultural products, organic agricultural products and products with GAPcertification were 4.6%, 1.5% and 0.8% respectively. Pesticides were not detected in pesticide-free agricultural products. Their detection levels were less than their maximum residue levels. Their estimated daily intakes ranged from 0.0003% to 0.04302% of their acceptable daily intakes, of which the values have no effect on human health.

• The Transactions of the Korean Institute of Electrical Engineers B
• /
• v.55 no.11
• /
• pp.572-576
• /
• 2006

As the increasing of capacity and technology of power facilities, rotating machines such as turbine generators and water turbines are getting higher at capacity but smaller in size. Thus the monitoring and diagnosis of generators for fault detection and protection has attracted intensive interest. Most of electrical faults of rotating machines appear in their windings. In case of an after-fault in high capacity rotating machines, the recovering cost is usually very expensive and additional time is necessary for returning in a normal situation. In this paper, the magnetic flux patterns in air-gap of a generator under various fault states as well as a normal state are simulated by a conventional FEM tool. These results are successfully applied to detection and diagnosis of the short-circuit condition in rotor windings of a high capacity generator.

• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.121.4-121
• /
• 2001

We propose an off-line methodology for detecting a faulty backup roll that generates eccentricity components, under the condition that the feeding velocity, equivalently the angular velocity of roll, is not constant. From a newly devised speed angle conversion algorithm, we transform all process data into those of a virtual process under a constant feeding speed. This indirectly way, we can apply a spectral analysis to the original process. In addition, we develop an online detection method of roll eccentricity based on newly designed PLG sensor. This PLG sensor is robust because of applying magnetic proximity sensnors and non-contact measurement method.

• Proceedings of the Korean Institute of IIIuminating and Electrical Installation Engineers Conference
• /
• 1999.11a
• /
• pp.127-131
• /
• 1999

In this paper, using a wide-band AE sensor with the frequency range from 100[kHz], the frequency spectra of AE signals generated from the corona discharges of the needle-plane electrode was analyzed to determine the proper ultrasonic sensor. We also examined the relationship between the magnitude of corona discharge and the magnitude of AE signals in peak-to-peak value under the application of 60[Hz] AC high-voltage in oil. From these results, the main frequency spectra of AE signals emitted from the corona discharges of the needle-plane gap were found to be 130[kHz] by the fast fourier transform. The magnitude of AE signals was proportional to the magnitude of corona discharge and discharge current pulse with increasing the applied voltages. Also the detection of corona discharge point location by AE signals was found to be possible by using two sensors.

• 한국연소학회:학술대회논문집
• /
• 2002.11a
• /
• pp.223-229
• /
• 2002

Spark plug ionization signal could be useful in an internal combustion engine as a feedback signal for combustion diagnostics such as misfire detection, knocking detection and lambda control, but the signal has high level of cyclic fluctuation in an internal combustion engine due to residual gas, pressure, temperature, mixture composition in the spark gap. Because of this reason it is very difficult to apply ion signal to commercial engine control. In this Study, a correlation between A/F and spark plug ionization signal was studied in a constant volume chamber. Constant volume chamber with gas phase fuel(Propane) has homogeneous fuel composition , no mixture flow, same pressure and temperature on each test. The results show that mean chemi-ion signal has the highest correlation with A/F and intial pressure change has on effect on the thermal-ion signal and not on chemi-ion signal.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2002.10a
• /
• pp.453-458
• /
• 2002

A recently developed Hilbert-Huang transform (HHT) technique is applied to the detection of the damage locations of bridge structures. The HHT may be used to identify the locations of damages which exhibit nonlinear and non-stationary behavior, since the instantaneous frequency characteristics of the measured signal can be analyzed by the HHT. Numerical simulations were conducted on two bridge systems with damages using controlled excitations with sweeping frequency. Nonlinear plastic model using a gap element is employed to model the behavior of the cracked elements in the numerical simulations. The results indicate that the HHT method can reasonably identify the damage locations based on a limited number of acceleration sensors. Experimental study has been 실so carried out on a steel frame to confirm the applicability of the HHT to detect a structural connection with loosened bolts.

• Journal of electromagnetic engineering and science
• /
• v.13 no.4
• /
• pp.214-223
• /
• 2013

This paper proposes a theoretical analysis of hidden device detection and a design of multiband circular polarization patch array antenna for non-linear junction detector system application. A good axial ratio of circular polarization patch antenna is realized by a new approach that employs inclined slots, two rectangular grooves and a truncated ground for the conventional antenna. A good axial ratio of the 1.5 dB lower is measured by having an asymmetric gap distance between the ground planes of the coplanar waveguide feeding structure. The common ground plane of the linear array has an optimum trapezoidal slot array to reduce the mutual coupling without increasing the distance between the radiators. The higher gain of about 1 dBi is realized by using the novel common ground structure. The measured return loss, gain, and axial ratio of the proposed single radiator, as well as the proposed array antennas, showed a good agreement with the simulated results.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.13 no.1
• /
• pp.74-84
• /
• 1988

This paper presents feature extraction technique for linear image using edge detection algorithms. The process of edge finding consists of determining edge magnitud and direction by convolution of an image with a number of edge masks, of thinning and ghresholding these edge magnitudes, of linking the edge elemtnts based on proximity ans orientation, and finally, of approximating the linked elements by piede-wise linear segmentss. These techniques are intened to be general and opplications to terminal detection and road recognition tasks are described. The presentation will be helpful to other researchers attempting to implement similar algorithms.