The Research And Implementation Of Liquid Level Detection Algorithm Based On Arm Platform

There are some shortcomings in traditional liquid level detection ways.Such as artificial and mechanical liquid level measurement detectoin's accuracy is too low; piezoelectric liquid level detection's result is very vulnerable by impurities; various kinds of sensor Liquid level detection's price is very high, and their application scope is narrow. Liquid level detection based on image analysising have lower cost, high precision safety and reliability, etc.Topic described in this article has obtained the support of City Technology Bure- au special industry fund. We implement a liquid level detection system using the combination between image processing technology and embedded technology, directing on the deficiencies of traditional liquid level detection means. The basic flow of liquid level detection based on image processing get the pixel's actuate size by system calibration, then retrieve liquid line through image preprocessing and edge detection, finally compute the liquid level. This process can calculate liquid level accurately, however the image noise, the offset of camera and background interference and so on cause the bias of result when the background is complex.This article advance the algorithms in different stages based on the basic principles of the liquid level detection algorithm and the actual issue of them: In the stage of Image Filtering, it bring up improved Adaptive Median Filtering. This algorithm using a imporved noise discrimination method base on the analysising of adaptive Median Filtering method, and it make different treatment on the different types of noise. In the stage of image matching,antirotation image matching algorithm is be raised for resolving the problem of camera 's moving.This algorithm is ground on gradient direction histogram and improved Hausdorff distance. In the image segmentation stage,in view of the environment complex and natural light of the circumstances, the use of edge detection can not partition the image effectively to get liquid level line, this artile put forward a comprehensive segmentation algorithm based on edge detection and color dividion. In the liquid level geometry detection stage, in view of the liquid level line is parabola when the container is canned , and RHT-3 parabola detection algorithm'computation is large, it is low efficiency. So this article raise improved RHT-3 algorithm using gradient the direction. It can reduce the amount of accumulation invalid points.The experiment indicated that the algorithm is more efficient than RHT-3.Finally, the transplant method of embedded Linux system and embedded GUI is be given, and detail information about the design of system software and hardware to apply the algorithm to the practicing system is be illustrated.The actual operation of the system aslo indicated the effectiveness of the improved algorithms.