Photoelectric Sensor Circuit and Image Segmentation Method for Radar System

Abstract

In this exploration, based on the principle and system parameters of laser three-dimensional (3D) radar imaging technology, the corresponding photoelectric sensor circuit scheme is formulated. The sense circuit of avalanche photon diode (APD) converts the signal through the transresistance amplifier circuit. Then, LMH6629 is selected as a precision amplifier with low input noise voltage and low input error current. The capacitance is used as a compensation element to compensate the phase. For the power supply scheme, choosing the mode of switching power supply and LDO to work together can improve the efficiency of power supply and reduce the output of current ripple. At the same time, semantic segmentation is carried out for the obtained photoelectric images. Based on the traditional spatial pyramid pooling algorithm, the fusion of mean intersection over union and cross information entropy loss function is introduced to improve the weight of local image region. In the experiment, Multisim software is used to simulate the circuit. The APD reverse bias voltage is set to 90 V, and the multiplication coefficient is 98.7. The feedback resistance, bandwidth, phase compensation capacitance and other parameters are further calculated. It is found that there is obvious self-excited phenomenon in the output waveform of the transresistance amplifier without phase compensation capacitor. When the feedback capacitance reaches 0.8 pF, the oscillation phenomenon is obviously reduced; further calculation shows that the bandwidth of transresistance amplifier is 230 MHz, and the noise of APD power supply is mainly caused by BUCK switching power supply switch when the bottom noise of oscilloscope is ignored. However, the noise is suppressed under the action of the back-end LDO device; after the loss function is introduced, the contour of the photoelectric image is preserved completely, and then the more accurate segmentation results are obtained.