Method for detection optimization in image-based decision systems
View Patent ↗A systematic way of linking structure constraints of components of a common object and checking and resolving their inconsistency is used to improve detection results in image-based decision systems. A multilevel Chain-And-Tree (CAT) model is used to direct processing using both forward and backward scans through the related components. Since components occur as parts of an object, the context (relational structure) in which the component appears can be used to reduce noise and variation affects. In the method, object knowledge is translated into constraints between components. The constraints are used to enhance feature detection, defect detection, and measurement accuracy and consistency.
1. A detection optimization method using a multilevel Chain-And-Tree model comprises the following steps:
(a) Performing a forward optimization scan initialization using the root node;
(b) Performing a forward optimization scan for each node using the detections and relations from its preceding nodes having at least one forward optimized result output;
(c) Performing a backward optimization scan initialization using the end node;
(d) Performing a backward optimization scan for each node using the detections and relations from its descending nodes having at Icast one backward optimized result output.
2. The method of claim 1 further comprises a defect detection step.
3. The method of claim 2 wherein the defect detection step includes gross defect detection.
4. The method of claim 2 wherein the defect detection step includes local defect detection.
5. The method of claim 1 wherein the forward optimization scan for each node comprises:
(a) Performing local detection for the current component having a current component raw data output;
(b) Performing an estimation using the previous component forward optimized raw data, the current component raw data and their relationship having a forward optimized current component object detection result output;
(c) Performing an update for the current component raw data using the forward optimized current component object detection result having a forward optimized current component raw data output.
6. The method of claim 5 wherein the estimation uses a structure-guided estimation method that optimizes a cost function.
7. The method of claim 5 wherein the update comprises error estimation and at least one adjustment step.
8. The method of claim 7 wherein the adjustment step performs no adjustment when the gross error is greater than a limit.
9. The method of claim 1 wherein the backward optimization scan for each node further comprises
(a) Performing an estimation using the previous component backward optimized raw data, the current component raw data and their relationship having a backward optimized current component object detection result output;
(b) Performing an update for the current component raw data using the backward optimized current component object detection result having a backward optimized current component raw data output.