|
<< 点击显示目录 >> 主页
mappVision帮助助手 > 机器视觉帮助 > mapp Vision > 用例 > 匹配 - 自学和后续对象搜索 匹配 - 对匹配等级值（分数）的影响 |
参数Score(n)是对找到的匹配结果与搜索所教模型匹配程度的评估。
以下参数对结果有影响。
•最小分数
•模型层数
•最大重叠
MinScore是一个最小评估值，它决定了找到的对象是否被视为有效匹配。顾名思义，Score和MinScore这两个参数并不直接相关或相互依赖，但MinScore对Score有影响，因为MinScore所接受的匹配的类型和质量可以相应地排除Score等级的候选对象。
ModelNumLevels（模型级数）决定了教入模型和所用图像以较低分辨率（关键词：金字塔级数）保留的频率。ModelNumLevel= 1 表示 mapp Vision 人机界面应用程序中显示的原始图像。每增加一个高于 1 的 ModelNumLevel，分辨率就降低一半。也就是说，如果ModelNumLevel= 3，原始图像的分辨率将降低四分之一。搜索算法从配置的最高级别（即最低分辨率）开始搜索，从ModelNumLevel= 1 开始的所有级别都必须至少与 MinScore 相对应。
搜索结果就是ModelNumLevel= 1 的分数。
注：这就是结果返回值远高于MinScore 设置值的原因。ModelNumLevels设置得越高，MinScore必须设置得越低（因为分辨率较低），但这并不意味着得分也低，因为得分始终是最低级别（原始图像）的值。
•MinScore= 0.7（对每个级别都有效） •模型级数= 3 |
•MinScore= 0.7（对每个级别都有效） •模型数量级 = 3 |
指定级别的分数结果示例 |
|
•第 3 级：候选者得分 0.7。接受。 •第 2 级：候选者得分 0.8。接受。 •第 1 级：考生得分为 0.9。接受。 |
•第 3 级：候选者得分为 0.5。弃权。 •第 2 级：候选者得分为 0.7。 •第 1 级：候选者得分为 0.75。 |
得分（n）的最终结果 |
|
最终结果为Score= 0.9（第 1 层的值） |
没有结果，因为第 3 层的结果已经太低。 可能的 解决方法 MinScore或减少ModelNumLevels。 |
信息：
ModelNumLevel 越高，分辨率越低。这将加快函数的运行速度，但在某些情况下，分辨率的降低会对模型产生很大的影响，以至于在执行过程中分数过低，无法提供最终结果。
信息：
在特殊情况下，根据教导过程自动计算出的 ModelNumLevel 可能会过高。这可能导致只能通过较低的 MinScore 找到非常好的候选模型（高分）。
Parameter Score(n) is an assessment of how well a found Matching result matches the model taught-in for the search.
The following parameters have an influence on the result.
MinScore is a minimum assessment value as to whether a found object is accepted as a valid match at all. Contrary to what the name suggests, the two parameters Score and MinScore are not directly linked to or directly dependent on each other, but MinScore has an influence on Score in that the type and quality of matches accepted with MinScore can accordingly exclude candidates for a Score grade.
ModelNumLevels determines how often the taught-in model and the used image are reserved in reduced resolution (keyword: pyramid levels). ModelNumLevel = 1 is the original image displayed in the mapp Vision HMI application. The resolution is reduced by half for each additional ModelNumLevel above 1. This means a quarter of the resolution of the original image if ModelNumLevel = 3. The search algorithm starts the search at the highest configured level (i.e. at the lowest resolution), and all levels down to ModelNumLevel = 1 must correspond to at least the MinScore.
The result is then the Score of ModelNumLevel = 1.
Note: This is the reason why the result comes back with a much higher value than is set with MinScore. The higher the ModelNumLevels are set, the lower the MinScore must be set (because of the lower resolution), but this does not mean that the score is also low since the score is always the value of the lowest level (original image).
•MinScore = 0.7 (valid for each level) •ModelNumLevel = 3 |
•MinScore = 0.7 (valid for each level) •ModelNumLevel = 3 |
Sample Score results for the specified levels |
|
•Level 3: Candidate has Score 0.7. It is accepted. •Level 2: Candidate has Score 0.8. It is accepted. •Level 1: Candidate has Score 0.9. It is accepted. |
•Level 3: Candidate has Score 0.5. It is discarded. •Level 2: Candidate has Score 0.7. •Level 1: Candidate has Score 0.75. |
Final result for Score(n) |
|
The final result is Score = 0.9 (the value of level 1) |
There is no result because the result in level 3 is already too low. Possible workaround: MinScore or reduce ModelNumLevels. |
Information:
The higher the ModelNumLevel, the lower the resolution. This makes the function faster, but in some cases the reduced resolution changes the model so much that it results in a score that is too low during execution and does not provide a final result.
Information:
In special cases, ModelNumLevel calculated automatically from the teach-in process can be too high. This can result in very good candidates (high score) being found only with a low MinScore.+++++++++