LED Display Panel Detecting System

The LED panel has many technical advantages, such as high brightness, long service life, low power consumption, stable performance. In recent years, it is becoming more and more popular as flat panel display mainstream product. An important development tendency of Chinese LED industry is standardization and normalization, which requires the detecting process of LED display manufacturers should be standardized.But in China, almost all the LED display manufacturers use eyes for the brightness detection of LED display, which affecting detection speed and quality detection. Therefore, this paper puts forward the LED display panel detection system based on the machine vision.This design set up the hardware structure of this system based on the structure of the machine vision system, using the LED control system to control LED display, using the camera to take photos of LED display unit, using acquisition card to transfer the data to the computer, using the digital image processing software to deal with the pictures and display the test results. This system simulates the method of artificial detection, using camera to take photos of black、scanning、red、green and blue pictures to finish the detection of all the problems of the testing unit board.The core of this system is the digital imaging processing software. The software has two core problems:light coordinate (a pixel in the area of a light area) and grey value collection of the lights area. Due to the "dark light" and "death light"、"scanning light position" cannot be directly positioned, we put forward "learning pictures" method for positioning the light coordinate of the detecting LED panel pictures. Learning pictures is the pictures of a type of LED panel displaying normally. First of all, we use median filtering to deal with the noises of learn pictures and detection LED panel pictures. And then we use the Zhang Zhengyou calibration method to recalibrate the radial distortion of the pictures. Through the implementation of the above methods, we reduce the testing error brought by the external environment and hardware equipment. Now we use OSTU method for learning images segmentation, use Freeman chain code represent each LED light outline to position each LED lamp, use the method of moment to calculate each center of mass of each LED lamp area. And then build the affine transformation relation between learning pictures and testing unit board pictures, which will transform the lights’coordinate of study pictures to the testing unit board pictures. After positioning the lamps coordinates of testing unit board, we use the Scan Line Seed Fill Algorithm to collect the gray value of the connected domain of each light, and calculate the average of that as the light grey value. We find the relation between luminous intensity(nit) to gray level of different type of LED panel,and then we get the standard that judge whether the light normal or not.Finally, this paper introduces the component of this system software and illustrates the operation method.