A real-time surface inspection system for precision steel balls based on machine vision

Abstract

Precision steel balls are one of the most fundament components for motion and power transmission parts and they are widely used in industrial machinery and the automotive industry. As precision balls are crucial for the quality of these products, there is an urgent need to develop a fast and robust system for inspecting defects of precision steel balls. In this paper, a real-time system for inspecting surface defects of precision steel balls is developed based on machine vision. The developed system integrates a dual-lighting system, an unfolding mechanism and inspection algorithms for real-time signal processing and defect detection. The developed system is tested under feeding speeds of 4 pcs s−1 with a detection rate of 99.94% and an error rate of 0.10%. The minimum detectable surface flaw area is 0.01 mm2, which meets the requirement for inspecting ISO grade 100 precision steel balls.

76‐5: Late‐News Paper: Enhanced Productivity Performance through Detecting Defects on Surface of AMOLED Display

“…Even similar product parts have differences in material and design, so the processor must be customized to these designated parts. In particular, the algorithm of the surface inspection program of AMOLED display should be designed by giving sufficient specification, tolerance and numerical guide line to the program developer, unlike the work guide to the operator, and this guide line should be fully recognized and operated by the machine [3]. It should be prepared by algorithm that classifies the defective location, image acquisition, noise analysis, image analysis and defect type of AVI test program products.…”

Application of Deep Learning Convolutional Neural Networks for Internal Tablet Defect Detection: High Accuracy, Throughput, and Adaptability

“…9,18 In addition, automatic inspection and defect detection based on deep learning CNNs have been widely adopted in many fields. 19 CNN methods have been shown to achieve high throughput quality control during the manufacture of metallic rails 20 and steel surfaces, 21 demonstrating their widespread adoption.…”

An End-To-End Model for Pipe Crack Three-Dimensional Visualization Based on a Cascade Neural Network

“…In addition, the traditional method relies heavily on the experience and knowledge of the algorithm designer, which leads to certain technical requirements for the detection staff, and hence, a low detection accuracy [3][4][5][6]. Some detection systems focus on the speed of image recognition [7,8], while others focus on accuracy [9,10]. Other algorithms analyze the direction and composition of cracks based on the information of the front and rear frames, and cannot implement end-to-end detection [11].…”