Search

In order to overcome the shortcomings of the measurement methods for gas cumulative flow, especially in the long-term continuous measurement, a novel memristor-based gas cumulative flow sensor is proposed in this paper. The sensor is mainly composed of a hot-wire resistor, the memristor module, a feedback compensation network and some basic circuit components. According to the thermal balance equation and the memory characteristics of memristor, the sensor can measure the gas cumulative flow by combining the memristor resistance with the cross-sectional area of the pipe. In order to verify the design of memristor-based gas accumulation flow sensor, experiments are carried out under a constant gas flow velocity, a gradually increasing flow velocity, and a randomly varying flow velocity. The active region thickness D of the selected TiO2/TiO2-n memristor is 40 nm. The hot-wire used is a tungsten wire having a radius of 10 μm and a length of 2 mm. The gas measured is air at a temperature of 300 K in the paper. The experimental results show that the sensor achieves the measurement of gas cumulative flow. When gas flow velocity at the center of the pipe is 5 m/s, the measuring range is approximately 0–20 m3 and the sensitivity within the effective range is approximately 28.44 Ω/m3. When the gas flow velocity, the memristor pairs increase and the gain of the amplifier changes, the measuring range, and the sensitivity of the sensor will changes accordingly. [ABSTRACT FROM AUTHOR]

In order to expand the measuring range of the exposure sensor and solve the current fluctuation caused by the memristor resistance, a novel exposure sensor based on reverse series memristor is proposed. The exposure sensor mainly consists of the photoresistor, the current limiting resistor and the reverse series memristor. The resistance of photoresistor varies with the illuminance, causing the loop current change. Based on the memory feature of the memristor, its resistance can be used to characterize the integral value of the illuminance for a period of time. Thus the measurement of the exposure can be realized by the exposure sensor based on reverse series memristor. In the exposure sensor, two series memristors constitute the reverse series memristor and they are connected in serial oppositely. And the sum of their normalized thicknesses is 1. Experimental results show that the exposure sensor based on reverse series memristor can measure the exposure and its measuring range is larger than other measurement methods. The measuring range is from 0 to 600 lx·s when the illuminance is 100 lx. With the increase of illuminance, the measuring range of the sensor will expand while the sensitivity will decrease. Because the resistance of the reverse series memristor almost remains unchanged during measuring exposure (the standard deviation of the change is only 0.43 Ω), the influence of the total resistance’ increase on the loop current can be avoided. Thus it can avoid the impact on the measurement of the exposure and reduce the error introduced by the change of the memristors resistance. As the number of memristors increases, the measuring range will expand and the sensitivity will improve, but the symmetry of the reverse series memristor will deteriorate. Hence the number of memristors can’t be excessively large. The variation tendency of the memristor resistance curve and the normalized thickness curve are consistent, and they rise along with the increase of the exposure. In addition, the measuring range of the exposure sensor will expand gradually while the sensitivity will decrease as the resistance of the current limiting resistor increases. [ABSTRACT FROM AUTHOR]

Highlights • An edge detection network architecture based on deep convolutional network is presented. • A novel loss function is proposed to balance positive and negative samples. • The fast learning model is discovered (only 5.4k iteration) for edge detection. • The better generalization ability is verified at different dataset. Abstract In this paper, we propose an edge detector based on feature re-extraction (FRE) of a deep convolutional neural network to effectively utilize features extracted from each stage, and design a new loss function. The proposed detector is mainly composed of three modules: backbone, side-output, and feature fusion. The backbone module provides preliminary feature extraction; the side-output module makes network architecture more robustly map features from different stages of the backbone network to edge-pixel space by applying residual learning, and the feature fusion module generates the edge map. Generalization ability on the same distribution is verified using the BSDS500 dataset, achieving optimal dataset scale (ODS) F-score = 0.804. Cross-distribution generalization ability is verified on the NYUDv2 dataset, achieving ODS F-score = 0.701. In addition, we find that freezing backbone network can significantly speed up training process, without much overall accuracy loss (ODS F-score of 0.791 after 5.4k iterations). [ABSTRACT FROM AUTHOR]