A system, method and model for identification of parents of a child found missing from the image of the child

Abstract

A missing child facial image is taken as input to the model. From the input image, facial image features are extracted using CNN network. These CNN feature arrays will be compared with the parent’s facial features similarity of the existing database. Further, computes the scores of similarities. The top ten similarity scores are used to extract their parent index images. In this model, a trained parent’s facial images database is used to assess the missing children’s parents using the child's facial image. Out of all assessed parent’s images the parents who have more facial similarities will be called and asked to identify their child. Further, one can open the mode of communication. This method and system can be used as a tool for identifying a missing child’s parents in the unfortunate event of a child getting missed in any congregation, gathering etc. It can help prevention of social evils etc.

Description

FORM 2THE PATENTS ACT, 1970[ 39 of 1970 ]&THE PATENTS RULES, 2003COMPLETE SPECIFICATIONA System, method and model for identification of parent of a child found missing fromthe image of the child( Section 10 ; Rule 13 )Applicants & Inventors(i) Dr. T MEENPAL(ii) Madhu Oruganti(iii)Ankita MeenpalAll Indian NationalsHaving address atAsst. Professor (Grade-I)(i) Department of ECE, NIT ( National Institute of Technology)Raipur, India-492010(ii) Research Scholar, Department of ECE, NIT Raipur, India-492010(iii)Research Scholar, Department of ECE, NIT Raipur, India-492010The following specification particularly describes the invention and the manner inwhich it is to be performed2INTRODUCTION:-This invention relates to the field of information technology. Still further, thisinvention relates to the field of image processing. Furthermore this invention relates toconventional neural networks. Still further, this invention relates tothe development of aninnovative method for parent identification from missing child facial images. Still moreparticularly, the invention relates to the development of an innovative method for parentidentification from missing child facial images.BACKGROUND OF INVENTION:-One of the biggest problem faced in country like India, is the case of missing children.Children, especially while going in congregations, fairs etc and in other crowded places areprone and vulnerable in getting missed from their parents or guardians.This incidents pose a grave threat to the society in general and causes anxieties amongparents in retrieving their dear children. Also, this causes high emotional stress and traumaamongst parents and relatives, siblings etc, as loss of a family member, temporarily though iscapable of causing severe shocks and would take longer time to heal, even after retrieval ofthe child. Also, from the Child’s point of view, it undergoes severe psychological stress andalso, faces potential threat of issues like child trafficking, abuse etc.Missing children from parents in various situations, like, child abduction by strangers,missing at public parks, bus stands etc is one of the challenging and emerging issues. If thechildren are not able to express their parent’s details, then it would be quite a challenging taskto address the problem. It is frightening that these millions of missing children end up inforced labour, slavery and sex work and sadly, this has turned into a lucrative business. Hereis a model to address the problem which gives the details of his / her parents.Due to huge growth in population, it is quite a challenging task to identify the parents ofmissing children in a manual manner until a smart model is introduced to address the problemwhich can be solved by fast improvement and in progressive manner. Verifying parents of3missing children is an emerging task if the process is much delayed it would impact on childlife.OBJECT OF THE INVENTION:-A primary object of the present invention is to create a model for identification of a child,who is missing from parents and lost in any congregation, gathering etc for identifying itsparents.Further, the above mentioned objective may be attained by updating the database of themissing child as well as of the parents based on the complaint received in police departments.Still further, it is an object of the present invention, that out of all the assessed parent’simages the parents who have more facial similarities will be called and asked to identify theirchild. Further, one can open the mode of communication with the concerned persons in thisinvention.Detailed Description of the Invention:-The proposed invention consists of two major units. One is extraction of child and parent’sfeatures using CNN network. Second one is computing similarity scores of parent andmissing child features. Based on these scores, the parents with highest scored features areconsidered or top ten scores can be considered for parent’s images.In pre-processing stage (101),A. Facial regions are extracted by removing background.B. Alignment of face by frontalization techniques.C. Lighting balance by histogram equalization technique.Facial feature extraction using CNN network (102),A. Output of 101 is the pre-processed image of missing child and is fed to featureextraction by using VGG16 network (102).B. The child facial features and parents image features are computed by using cosine4similarity assessments (103).C. Based on the cosine similarity scores, highest cosine similarity scored parents areconsidered as parents of missing children.In pre-processing 101 stage, the considered missing child image / photo is transformed intorequired dimensions like, resize operation, face normalization, etc. In facial feature extraction102 stage, the processed facial image gets transformed by the VGG16 network which is atrained model. In this stage from processed input image, facial features (N number offeatures) like edges and shapes of facial parts are extracted from facial image.In the 104 stage parent image dataset which are VGG16 network based features of allcollected parent images with size of M x N features. These features of each parent image arecompared with each missing child feature in correlation analysis 103 stage. In this stage,similarity scores are computed for all the parent images features with respect to missing childfeatures. Based on these scores, the highest scored parent is confirmed as the parent of themissing child.The Proposed model diagram consists of five blocks. Main idea of the proposed model is toidentify the parent facial image based on missing child facial image. Our model works onextracting facial similarities of blood related people like father-daughter, mother-daughterfather-son, mother-son.The missing child’s facial image is taken with an HD camera and pre-processed in stage 101.The parent’s facial image is collected from the respective police departments or child welfaresocieties who complained about their missing child. The collected trained facial images ofparents with the size of M x N features are computed using VGG16 pre-trained model instage 104. Where M is the total number of parent images and N is the number of VGG16features. The child input image Ich is p X q X c size is sent to theVGG16 feature extractionblock as shown in Fig.2. The output of these VGG16 features is sent to the correlations5analysis stage 103.6

Claims

CLAIMSWe Claim1) This model consisting of pre-processing stage, wherein, facial regions are extracted byremoving background, followed by alignment of face by frontalization techniques and furtherby lighting balance by histogram equalization technique. In the facial feature extraction stage,wherein output of 101 is the pre-processed image of missing child and is fed to featureextraction by using VGG16 network (102) and the child facial features and parents imagefeatures are computed by using cosine similarity assessments (103) and based on the cosinesimilarity scores, highest cosine similarity scored parents are considered as parents of missingchildren2) The pre processing stage as claimed in claim 1, wherein the pre-processing 101 stage, theconsidered missing child image / photo is transformed into required dimensions like, resizeoperation, face normalization, etc. In facial feature extraction 102 stage, the processed facialimage gets transformed by the VGG16 network which is a trained model. In this stage fromprocessed input image, facial features (N number of features) like edges and shapes of facialparts are extracted from facial image.3) The model for parent identification as claimed in claim 1, wherein the 104 stage parentimage dataset which are VGG16 network based features of all collected parent images withsize of M x N features. These features of each parent image are compared with each missingchild feature in correlation analysis 103 stage. In this stage, similarity scores are computedfor all the parent images features with respect to missing child features. Based on thesescores, the highest scored parent is confirmed as the parent of the missing child.4) A method for identification of parents of a child which has gone missing is given, whereinthe missing child’s facial image is taken with an HD camera and pre-processed in stage 101.The parent’s facial image is collected from the respective police departments or child welfaresocieties who complained about their missing child. The collected trained facial images of7parents with the size of M x N features are computed using VGG16 pre-trained model instage 104. Where M is the total number of parent images and N is the number of VGG16features. The child input image Ich is p X q X c size is sent to theVGG16 feature extractionblock as shown in Fig.

2. The output of these VGG16 features is sent to the correlationsanalysis stage 103