Research On Monitoring System Of Underground Sump Level Based On Visual Recognition

With the development of the coal industry,coal mine safety has always been the focus of most people's attention.Among them,the problem of excessive groundwater overflow has not been solved.During the coal mining process,the gushing groundwater collects in the small water tank of the roadway,and the liquid level in the small water tank is detected by the liquid level meter.The level gauge has also evolved from the initial contact to a non-contact level gauge.However,in actual production,the pressure type liquid level gauge will have a large measurement error due to the blockage of impurities such as cinders and vermiculite in the water tank,and the reliability is low.Traditional non-contact liquid level gauges such as ultrasonic and laser level gauges can avoid these problems,but are limited by downhole power,and since both liquid level gauges are measured by the principle of liquid level reflection,the liquid level fluctuations The influence of interference is huge,so the traditional non-contact liquid level gauge still has certain limitations.This paper designs a liquid level monitoring system based on visual recognition technology through visual monitoring method combined with image processing technology.The system consists of ultrasonic sensor,visual recognition liquid level meter and water level controller.In terms of hardware,STM32F103ZET6 is selected as the CPU of the visual recognition liquid level meter,and with the ATK-OV7725 camera with its own FIFO chip,the camera can allow the CPU to read image information at any speed to meet the requirements of image processing.After the camera position calibration,Collect image information of small water bins in mine roadway;some important peripheral circuit designs include: power supply circuit based on flyback transformer,ultrasonic transmitting and receiving circuit,reset circuit and communication circuit based on MAX811 S chip,etc.;hardware of water level controller The circuit design includes leakage blocking,current and voltage detection circuits,and the like.In terms of software,the overall architecture of the system software was designed,and the ?C/OS-II embedded operating system was transplanted.Based on the operating system,the application programming(IAP),communication program,data acquisition program and fault subroutine were designed.The liquid level recognition algorithm is designed.After the image captured by the camera is processed by graying,binarization,median filtering and edge detection,the liquid level data is calculated through the pixel point information,and the liquid level height is visually recognized.When the height of the liquid level detected by the ultrasonic sensor is consistent,the current liquid level is obtained.The simplification and optimization method of the convolutional neural network algorithm in the embedded system is discussed.The visual recognition level meter communicates with the water level controller through RS-485.The water level controller can start and shut down the drain pump according to the liquid level information transmitted by the liquid level meter.It has leakage lockout,overvoltage,undervoltage,overload,short circuit and lack of The equal protection function protects the motor during condition monitoring of the drain pump.The liquid level monitoring system has low cost,small volume and high stability,and has the function of remote online download and upgrade program.In the laboratory simulation environment,the recognition rate of the static liquid level is as high as 100%,and the average value of the single recognition time is 3.24 S.The recognition rate of the wave surface is 91.67%,and the average value of the single recognition time is 4.56 S.It is suitable for monitoring the liquid level of small water tanks with small fluctuations.It has high recognition accuracy and fast speed,and meets the needs of automatic monitoring of liquid level in small water tanks of coal mine roadways.