Suspect face retrieval system using multicriteria decision process and deep learning

Suspect face retrieval system using multicriteria decision process and deep learning

---

연구 분야: Strategies

---

---

학회: Multimedia Tools and Applications

---

초록

The identification and apprehending of suspects by law enforcement authorities rely heavily on facial sketches. The sketch artist creates sketches based on the witnesses’ memories. Sketch artists are few and limited in their availability. It is also evident that as time passes, the eyewitness forgets many of the important details, which can be expensive in time-sensitive investigations. The sketch was used to obtain the suspect’s image through the state-of-the-art sketch-photo retrieval model, which missed the relevance of time sensitivity. A linguistic description-based suspect face image retrieval approach is presented in this study. In the proposed approach, the facial attribute-value pair is extracted from eyewitness descriptions. Facial attribute saliency is also studied in this work and validated with the Fuzzy Analytic Hierarchy Process (FAHP) model. A weighted score is computed to retrieve the suspect face images. The effectiveness of the proposed method is assessed by comparing it to existing linguistic sketch-based retrieval methods as well as the sketch to photo retrieval models. As compared to state-of-the-art approaches, experimental results give an accuracy of 94.98%.

---