2019 年 70 巻 2E 号 p. 124-135
This paper focuses on the development of a work analysis support system for distribution processing and presents a case study of work analysis at a retail clothing order fulfillment center. The system comprises ultrasonic sensors for measuring a worker's ow line and a smartphone for measuring the worker's dominant hand acceleration. Models for estimating the worker's motion were derived from the data obtained. Candidate estimation models were the statistical models linear discriminant function, decision tree, and k-nearest neighbor algorithm, and we used generalization error by cross-validation as an index for choosing the optimal estimation model. The system was applied to the example of analyzing order picking in a clothing distribution center. The κ coefficient obtained indicated the degree of matching between the results of video analysis and tracking by our system to be 0.6 or more, which conffiermed the validity of our system.
