Multi-level smartphone screen defect detection method
A smart phone and defect detection technology, applied in computer parts, image data processing, biological neural network models, etc., can solve the problems of not meeting the independent and identical distribution conditions, uneven data set quality, and low discrimination of defect samples. , to achieve the effect of reducing the input of human labeling, improving the detection ability and improving the efficiency and real-time performance.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0051] A multi-level smart phone screen defect detection method, the overall process is as follows figure 1 shown, including:
[0052] Step 1, such as figure 2 , manually mark the defects in the image, and then separate the foreground defect image from the background image; on the basis of the marked foreground defect image, enhance the color and size of the foreground defect image (such as image 3 shown), to expand the number of samples and morphological diversity of foreground defects; finally, combine the enhanced foreground defect images and background images to generate a data set with reliable labels and balanced categories, suitable for actual production scenarios; defects include hair, dirt, black Spots, scratches and discoloration;
[0053] Step 1.1. Manually mark the location and category information of the rectangular area where each foreground defect image is located in a small number of images, and then separate the foreground defect image and the background i...
Embodiment 2
[0086] On the basis of Example 1, the multi-level convolutional neural network model training and testing is carried out by using the data set generated in step 1. The experimental data includes 5422 pieces of real scene annotation data, and the experimental evaluation indicators are defect detection accuracy, recall Rate and F1 score, the accuracy rate (precision) indicates how many samples in the predicted results are correct, and the recall rate (recall) indicates how many positive samples in the predicted results are correctly detected. F1 is defined as follows:
[0087]
[0088] The definition of correct detection is that the IoU (overlap) between the detected area and the marked area is greater than or equal to 0.5, and the detected type is consistent with the marked type. The IoU calculation is the ratio of the area of the intersection and union of the "predicted area" and the "real area". The experimental results are shown in Table 1 below.
[0089] Table 1 Defec...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More