Abnormal image detection method and device, equipment and medium
An abnormal image and detection method technology, applied in the field of image processing, can solve the problem of low accuracy of portrait clustering, and achieve the effect of improving the accuracy.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0022] figure 1 A schematic diagram of the abnormal image detection process provided in the embodiment of the present application, the process specifically includes the following steps:
[0023] S101: For each pre-built isolated tree, obtain the image feature and the first feature value corresponding to the first root node in the isolated tree, and obtain the second feature value of the image to be detected corresponding to the image feature, if all The first root node includes child nodes, and the length of the first path of the image in the isolated tree is updated, and according to the size of the first eigenvalue and the second eigenvalue, the corresponding value of the image is determined. the first target child node of the first root node, and update the first target child node to the first root node, and continue to determine whether to update the first path length of the image in the isolated tree, if the If the first root node does not contain child nodes, it is dete...
Embodiment 2
[0038] In order to further improve the accuracy of portrait clustering, on the basis of the above-mentioned embodiments, in the embodiments of the present application, the process of constructing a preset number of each isolated tree includes:
[0039] Obtain the feature value of each preset image feature of each sample image in the target sample set;
[0040] Create the second root node of the isolation tree to be constructed, randomly select any preset image feature as the first target feature, randomly select any feature value of the first target feature as the first target feature value, corresponding to the second The root node saves the first target feature and the first target feature value; for each sample image in the target sample set, according to the feature value of the first target feature and the first target feature of the sample image value, determine the sample image corresponding to the first child node or second child node of the second root node; for any c...
Embodiment 3
[0064] In order to further ensure the accuracy of the constructed isolation tree, on the basis of the above embodiments, in the embodiment of the present application, before the creation of the second root node of the isolation tree to be constructed, the method further includes:
[0065] Counting whether the number of sample images contained in the target sample set is greater than a preset number threshold;
[0066] If not, the subsequent step of creating the second root node of the isolated tree to be constructed is performed.
[0067] At present, most anomaly detection methods hope to have more sample images to complete the abnormal image detection more accurately, but in the embodiment of the present application, using a target sample set with fewer sample images can often achieve better results . A large number of sample images will reduce the ability of the embodiment of the present application to isolate abnormal images, because normal images will interfere with the p...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More