An Optimized SSD Target Detection Algorithm Based on K-Means Clustering

In response to the problem that the default box size and shape of the SSD network model need to be manually set based on experience and the lack of specificity for different data, this paper uses the k-means clustering method to optimize the default box setting method of the SSD network to make the default box more consistent with the data, enhancing the self-adaptive ability of SSD default box positioning regression, thereby improving detection accuracy and detection speed. The algorithm is applied to actual aluminum defect detection, the defect detection accuracy reaches 77.6% mAP, which is 2.86% higher than the original SSD512 model, and the detection speed is increased from 37 FPS to 39 FPS.