1 / 7

15-463: Computational Photography- Gait-Based Human Recognition

Girish Jattani. 15-463: Computational Photography- Gait-Based Human Recognition. Motivation/Prior Work. Automatic Gait Recognition Employs statistical analysis of human walking patterns to identify humans in a scene Motion can be simplified to three segments

chandler
Download Presentation

15-463: Computational Photography- Gait-Based Human Recognition

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Girish Jattani 15-463: Computational Photography-Gait-Based Human Recognition

  2. Motivation/Prior Work • Automatic Gait Recognition • Employs statistical analysis of human walking patterns to identify humans in a scene • Motion can be simplified to three segments • Relevant center of pixel mass for each segment can be used to determine each position relative to the other • Variance can identify particular characteristics about each segment • The legs have high variance because they are usually spread apart • The torso has low variance because it occupies almost all of it's segment • The head has little variance in the middle of it's box

  3. Overall Algorithm • Perform image subtraction for each frame in the sequence with the first frame as the reference • Work in the greyscale space • Use Otsu's thresholding algorithm to adjust for contrast differences between frames • Perform morphological filtering to fill in holes from similar colored background and foregrounds • Find all potential regions of connectivity using 8-way connectivity • Divide each region ('blob') into three equally-sized segments corresponding to the head, torso, and the legs • Perform statistical analysis

  4. Results • First row: successes • Second row: failures

  5. Results (continued)

  6. Results (continued) • Contrast plays a huge role in the detection, to the point where slight variations in camera positioning greatly affect results:

  7. Issues • Variation of contrast was too difficult to overcome even with adaptive • Structural elements used for morphological processing need to be more specific to image • Occlusion is still not fully handled using gait based analyses • Slight camera movement greatly impacts results

More Related