A Novel Method for Handling Partial Occlusion on Person Re-identification using Partial Siamese Network

Abstract

Person-reidentification (Re-ID) is one of the tasks in CCTV-based surveillance system for verifying whether two detected objects are the same person or not. Re-ID visually matching one person or group in various situations obtained from different cameras or on the same camera but at different times. This method replaces the task of surveillance through surveillance cameras that was previously carried out conventionally by humans because it is prone to errors. The challenge of Re-ID is the pose of varied objects, occlusions, and the appearance of people who tend to be similar. Occlusion issues receive special attention since the performance of Re-ID can decrease due to partial occlusion. This can occur because the re-identification process relies on features of the person such as the color and pattern of clothing. The occlusion resulted in the feature not being caught by the camera resulting in a re-identification error. This paper proposed to overcome this problem by dividing the image into several parts (partial) and then processed in different neural network (NN) but with the same architecture. The research conducted is applying the CNN algorithm with the Siamese network architecture and applying the contrastive loss function to calculate the similarity distance between a pair of images. The test results show that the partial process obtained an accuracy of 86%, 77%, 68%, and 56% for occlusion data of 20%, 40%, 60%, and 80%. This accuracy is three to five percent higher than images without partial.