Improvement of recognition rate of towel mark recognition system

Flexible objects such as clothes are hard to be handled by robots. But the working conditions for laundry factories are severe because of dusts, heats and steams, and full automation systems which can handle clothes are highly required. To automate these factories, a cloth handling robot system has been developed. It can handle face towels and hand towels, successfully. But to handle various kinds of towels, it is required to recognize the marks on them. The marks on towels are weaved in cloth and it requires the special lightings. The image of the towel marks contains many texture noises, and it is generally hard to recognize the mark in many textures. Here, to recognize the marks, the inclined lighting system and the recognition method using HOG features have been introduced. Evaluating the HOG feature space distances makes robust recognition of marks on cloth textures. The system have embedded to the cloth handling system and proved its efficiency. But the system still has a little failures of recognition. To meet with the failures, here, the cause of the failure has been analysed, and the system has been improved by developing 3D wrinkles detection method.