A target detection method, device, equipment and readable storage medium
By performing density detection and adaptive block cropping on ultra-large pixel images, combined with background modeling and an adapted target detector, the problem of low detection accuracy in ultra-large pixel images is solved, and efficient target detection and tracking are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INSPUR SUZHOU INTELLIGENT TECH CO LTD
- Filing Date
- 2022-06-30
- Publication Date
- 2026-05-26
AI Technical Summary
Existing object detection algorithms cannot effectively handle ultra-large pixel images, resulting in low detection accuracy. Furthermore, the existing sliding window image slicing and stitching method further reduces the accuracy of object detection.
Density detection technology is used to dynamically and adaptively segment ultra-large pixel images. Image blocks are cropped according to the density detection results, and a suitable target detector is selected for detection. Background modeling is combined to remove non-target areas, thus optimizing the image segmentation and detection process.
It improves the accuracy and speed of target detection, reduces hardware requirements, enhances the cost-effectiveness of the target detection system, and supports target tracking applications.
Smart Images

Figure CN115131707B_ABST