Research On Fabrication Method Of Wearable Environmental Sensor Based On Visual Image Processing

In this paper,a method for preparing a sensor system combined with image processing is proposed,and a wearable pressure sensor and a wearable ultraviolet sensor are prepared based on this method.Traditional measurement systems are based on electrical methods,which convert external signals,such as pulse,ultraviolet light intensity,respiratory intensity,and other analog signals into electrical signals for characterization.Then the data is collected and processed by the data acquisition system.Based on these principles,sensors we built require complicated wire connections and complicated manufacturing processes.Moreover,capacitor-based sensors are susceptible to interference and piezoresistive sensors have drifting problems that need to be resolved.The non-contact preparation method of sensor measurement system based on visual image processing proposed in this paper can avoid complicated wire connections,thereby simplifying the preparation method and increasing the stability of the sensor measurement system.In our research,a flexible device based on visual signals is proposed.Structurally,this device is an integrated sensor with a transparent substrate and a black micropillar array adhered to it.In the pressure range of 0-3 kPa,this sensor has excellent pressure sensing performance,specifically,its pressure sensitivity is 0.133/kPa.Practically,the sensor can be used in outline obtainment of stuffs,especially,braille.Not only that,this device is very promising in the application of wearable devices,such as arterial pulse signal measurement.Combine the method of deep learning image sensor and ultraviolet sensor to realize the development of ultraviolet sensor system.Firstly,the deep learning classification method of convolutional neural network is combined with the ultraviolet sensor to realize the discrete measurement of the ultraviolet sensor in the training set.The limit of detection of the sensor was about 9 ?W/cm2 and the recognition rate of the network is above 90%under different environment light condition.The CNN Network test time only needs 3 s.Finally,an UV intensity recognition APP based on mobile convolutional neural network was developed.Because the deep learning classification method cannot realize the measurement of ultraviolet that does not exist in the training set,it cannot obtain a continuous sensor performance curve.Then the deep learning regression method is considered to make an ultraviolet measurement system.A UV sensing system based on this C3E estimation method was presented.The sensor after algorithm optimization has a sensitivity of 1 ?W/cm2.And the limit of detection is 9?W/cm2.The network test time is close to 3 s,and the error rate is 5%,which is very suitable to transplant to the embedded platform and provides solutions for compact instruments.