An FPGA-implemented method for real-time multi-dimensional feature extraction of sequence image targets
-
摘要:
目标区域多维特征的建模表征是实现目标跟踪搜索的前提。传统的特征提取先连通出目标区域,再计算目标特征,实时性不强。提出了一种基于像素连通域标记及目标特征同步计算的方法,目标区域连通完成即可输出目标特征。该方法建立了特征传递机制,在扫描图像时同步建立标号表、标号映射表与特征属性表,利用标号映射表关联了标号表与特征属性表。区域邻接时进行标号合并和特征属性同步传递计算,保证了目标特征提取的实时性。给出一种基于FPGA硬件设计的实现方案。仿真测试结果显示,此方法具有几项突出的特点:连通域标记耗时接近理论最小值;利用循环缓存区存储图像,资源消耗小;标记与计算并行流水处理,提升检测跟踪效率;多目标特征经测试验证结果准确,可有效支持后续的目标跟踪检测;具有理论与实用价值。
Abstract:The prerequisite of target detection and tracking is to model and represent the target based on multi-dimensional features extracted from the target region. The traditional target feature extraction needs to connect the target region first and then must calculate the target feature, which has still room for improvement in real-time. An advantage of the new method based on the synchronous calculation of pixel-connected domain markers and target features is that the target features can be output when the target region is connected. The process establishes a feature transfer mechanism, creates a marker table, a marker mapping table, and a feature attribute table simultaneously when scanning images, and uses the marker mapping table to associate the marker table and the feature attribute table. The marker merging and feature attribute transfer calculation are performed synchronously at the time of region adjacency, which ensures the real-time target feature extraction. As a result, the method proved to be effective because of an implementation scheme based on an FPGA system design. Simulation test results show that this method has several outstanding features: the time of connected domain markers is close to the theoretical minimum; the storage of images utilizes the circular buffer with low resource consumption; marking and computation are processed in parallel flow to improve detection and tracking efficiency; multi-target features are tested and verified to be accurate, which can effectively support subsequent target tracking detection; and it has theoretical and practical values.
-
Key words:
- target feature extraction /
- real-time /
- connected domain marker /
- image processing
-

计量
- 文章访问数: 20
- HTML全文浏览量: 1
- PDF下载量: 7
- 被引次数: 0