贝叶斯最小错误率在水下焊缝识别中的应用

Application of minimum error probability of Bayes was in underwater V-groove weld images processing

利用贝叶斯最小概率决策寻找V形焊缝左右边的数据点,即以一个特征观测量x1表示从左向右连续的高度位置逐渐降低的25个以上的白点,满足这一特征数据的概率大约为0.3,且以w1表示V形焊缝左边数据的类。连续扫描高度为20个像素、宽度为30个像素的小区域,为简单起见,以该区域内获得的白点数除以30作为x1的类条件概率密度,通过计算x1关于w1的后验概率,如果大于0.8,则把这些点归为V形焊缝的左边这一类别,从而对左边采用最小二乘法进行拟合,同理对V形焊缝右边的数据进行识别,并进行拟合。

The data dots which represented the Vgroove weld was obtained by the minimum probability decisionmaking of Bayes,the method was like this: x1 was a special metrical value which represented much more than 25 white dots whose heights were descendant little by little when the program scanned from the left to the right,and the probability of the condition was about 0.3 when detected by the program,and w1 represented the left of a V-groove weld.The program scanned a small rectangle region whose height was 20 pixels and width was 30 pixels,for simplicity,the number of the white dots lying in the small rectangle was divided by 30,and the consequence was considered as the value of the ex post facto probability,and if the value was bigger than 0.8,then,these white dots would be regarded as the left dots of the V-groove weld,at last,the left of it was fitted with these dots through the Least Square Method,the right of the V-groove weld could be detected and fitted as the same way as well.