一种立体视觉反馈强化训练模型

A stereopsis visual feedback strengthen training model

目前存在的立体视觉训练方法普遍缺乏反馈强化训练机制,针对这个问题,为了提供一种反馈的强化训练效果,给出了一种反馈强化训练模型及软件实现。该软件以随机点立体图(RDS)为训练视标,通过一次预训练来获得初始数据,在训练过程中以识别的正确率和反应时间为判断标准,找出双眼不易识别的RDS参数区间并进行强化训练,实现了反馈强化机制。测试结果表明,所得数据及解释符合能量模型,用此方法来训练立体视觉是切实可行的。

At present, the generally stereopsis training methods lack of a feedback mechanism, aiming at this problem, in order to provide a feedback strengthen training purpose, a feedback training model was given and the soft was also realized. Random dot stereograms（RDS） were used for the identification target. The first data was gotten by pre-training. Using the correct rate and reaction time for the judgment, the parameter range was searched that eyes distinguish RDS hardly, and then strengthen training was done on this parameter range. The feedback mechanism was realized. The test results indicate that the data is in line with energy model, and it is practical and feasible that using this way to train the stereopsis.