Precision shrapnel shrapnel shrapnel online quality detection system/shrapnel is laid underground track a large amount of used parts, to ensure the quality of the shell, need mechanical size of on-line detection of workpiece.
Machine vision inspection technology has a non-contact, on-line real-time, the appropriate speed and precision, the scene has advantages such as strong anti-jamming capability, to adapt to the progress and development of modern manufacturing requirements, and will have broad prospect of application in practice.
Combined with machine vision and image processing technology, designed and implemented a set of on-line detection system for quality of shrapnel.
The realization of the system, implementing 1 the following several aspects of research work.
In the environment and the actual testing demand of full research on the basis of the design of image acquisition system;
In order to ensure the dimension accuracy of measurement, focuses on the camera calibration technique;
3, establish proper image processing algorithms, realize the online quality detection of shrapnel, achieve the desired accuracy requirements.
In order to make the machine vision system to achieve the requirement of the precision and speed, and analyses the existing calibration methods, the principle of ant colony algorithm, neural network to do a simple introduction, puts forward a random ant colony algorithm to optimize the BP neural network used for camera calibration, compared with other calibration technology has improved in terms of accuracy and robustness.
Threshold segmentation, edge detection algorithms are briefly discussed, and select the most suitable solution, according to the characteristics of the shrapnel in the image acquisition system in the industrial camera lens with filter, greatly improve the quality of the collected image, and ultimately successful extraction shrapnel feature points.
The actual measure the quality of shrapnel, accuracy is 0.
1 mm, comply with the design requirements, and the results of measurement are discussed and the error analysis.
Research and system using results show that the detection system scheme and adopted by the method is feasible and correct, various performance and indicators have reached the expected requirement.