Walnut defect detection based on improved ELM and computer vision

Abstract

Objective: To address the issues of low accuracy and poor efficiency in walnut defect detection among existing food production enterprises. Methods: Proposed a fast non-destructive detection method for walnut defects that combined improved extreme learning machines and computer vision. Collected most of the surface image information of walnuts through computer vision, preprocess the image through Gaussian filtering, optimize color and texture features through iterative and information preserving variable methods, finally, by improving the butterfly algorithm to optimize the parameters of the Extreme Learning Machine （random weights and deviations）, fast non-destructive detection of walnut defects could be achieved, and verify the performance of the proposed defect detection method. Results: The experimental method could effectively discriminate various defects in walnuts. Compared with conventional methods, the experimental method had superior detection accuracy and efficiency in walnut defect detection, with a detection accuracy rate > 98.00% and an average detection time < 9.00 ms. Conclusion: Combining intelligent algorithms with machine vision technology can achieve rapid non-destructive detection of walnut defects.

Keywords

• food production
• walnut defects
• computer vision
• extreme learning machine
• butterfly optimization algorithm
• rapid non-destructive testing