Glass bottle crack detection method based on machine learning

• Main Authors: ZHANG TAO, DING BIYUN, LIU GANJUN
• Format: Patent
• Language: Chinese; English
• Published: 05.01.2021
• Subjects: ACOUSTICS; CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIRCHEMICAL OR PHYSICAL PROPERTIES; MEASURING; MUSICAL INSTRUMENTS; PHYSICS; SPEECH ANALYSIS OR SYNTHESIS; SPEECH OR AUDIO CODING OR DECODING; SPEECH OR VOICE PROCESSING; SPEECH RECOGNITION; TESTING

The invention discloses a glass bottle crack detection method based on machine learning. The method comprises the following steps: acquiring a sound signal generated when a metal rod knocks a to-be-detected glass bottle body by adopting a pickup; carrying out feature extraction on the collected sound signals, and extracting traditional features, short-time features and time-frequency features of the sound signals to obtain an initial feature set; performing feature selection on the initial feature set by adopting a shuffled leapfrog algorithm to obtain an optimal feature subset; taking the optimal feature subset as input of BPNN, carrying out BPNN training to obtain model parameters, then judging whether the glass bottle has cracks or not according to output of the BPNN, and finally obtaining a crack detection result; and according to the crack detection result, rejecting the glass bottles with cracks by adopting rejecting equipment. The glass bottle crack detection method provided bythe invention has the cha