Author Affiliations
1College of Information Science and Engineering, Changsha Normal University, Changsha, Hunan 410100, China2Physical Science and Electronics, Central South University, Changsha, Hunan 410083, Chinashow less
Fig. 1. Neighborhood method
Fig. 2. Specific method for selecting specific search area
Fig. 3. Tracking results of target tracking algorithm based on similarity feature estimation
Fig. 4. Performance comparison of different target tracking algorithms, under the experimental scene of enlarged target size
Fig. 5. Performance comparison of different target tracking algorithms, under the experimental scene of reduced target size
Fig. 6. SNR comparison of output results with different target tracking algorithms
No. | Configurationtype | Specificconfiguration | Configurationdata |
---|
1 | Hardware part | CPU | Intel Core 2 | Frequency/GHz | 1.2 | RAM/G | 2 | 2 | Software part | Operating system | XP | Development tools | 2015 VS platform |
|
Table 1. Specific configuration data of the experimental environment
No. | Configuration type | Configuration data |
---|
1 | Coding class | High class | 2 | Image tracking size | 4CIF、CIF | 3 | Encoding bandwidth/k | 2048、1024、768、384 | 4 | Coding format | 4MPEG | 5 | Reference code | XVID reference software | 6 | Code specific bit rate | 1024 | 7 | Encode the specificframe rate/(frame/s) | 25 |
|
Table 2. Specific configuration data of video coding in the experiment
No. | Configuration type | Configuration data |
---|
1 | Color space configuration | HSV space | 2 | Component configuration | H component | 3 | Quantization treatment series | 32 | 4 | Kernel function | Epanachnekov |
|
Table 3. Parameter specific configuration data
Scenario | Enlarge the target sizeexperiment scene | Reduce the target size of theexperimental scene |
---|
Track target number | 1 | 3 | 5 | 1 | 3 | 5 | Target tracking algorithm based onsalient features/ns | 623 | 789 | 817 | 626 | 749 | 860 | Algorithm based on fusionfeatures/ns | 571 | 577 | 634 | 524 | 581 | 612 | Algorithm based on backgroundweighting/ns | 597 | 682 | 701 | 582 | 654 | 693 | Algorithm based on convolutionalnetwork/ns | 412 | 421 | 430 | 461 | 487 | 496 | Target tracking algorithm based on kernelcorrelation filtering of dual feature model/ns | 532 | 597 | 631 | 624 | 636 | 675 | Algorithm based on similarityfeature estimation/ns | 245 | 344 | 435 | 267 | 355 | 430 |
|
Table 4. Time-consuming comparison of target tracking process with different algorithms