On 14th July 2015, a boy from Handia in Uttar Pradesh disappeared from his house. His family approached the police and a missing-person case was registered. On 23rd July, nine days after the boy disappeared, Goalpara Police found him more than 1,000 km away in Assam. He was admitted to a local child welfare centre. While the child was found, no one knew that he was the boy whose family was looking for him in UP.
The boy had autism. He remained at the child home in Assam for the next five years. In 2020, Telangana Police ran photographs through DARPAN, its facial-recognition system, and matched his photograph with the missing-child record from UP. Telangana Police informed Handia Police and his parents then travelled to Assam. They identified the child and the family was finally reunited. But it took five years and the child was 13 by then.
According to Telangana Police, DARPAN was being used to match photographs of missing children with children housed in child homes across India. The missing-child photographs came from records including FIRs, CCTNS and TrackChild. The photographs of found children were obtained from child homes and child-welfare authorities. At the time of the UP-Assam reunion, Telangana Police stated that DARPAN had helped trace and reunite 23 children.
The outcome of this story is not that the child was reunited with the family, but the fact that while he was found within nine days, it took five years for his family in UP to finally find him in Assam. For five years, a police station had a missing child on its records while a child-care institution had the child himself. What remained missing was the connection between the two. This is where the technology part of India’s missing-children story begins.
In the previous article, we discussed the system that gets activated after a child disappears, from the FIR and CCTNS to TrackChild, Khoya-Paya, railway rescue operations, Child Welfare Committees and eventual restoration. In this article, the question is different. Once the system has collected a photograph, CCTV footage, a shelter record or another digital trace, can technology connect information sitting hundreds or thousands of kilometres apart?
Over time, the child-protection database itself has changed. The Ministry of Women and Child Development has brought the earlier TrackChild portal for missing and found children and the Khoya-Paya application for missing and sighted children into the unified Mission Vatsalya Portal.
The TrackChild component remains integrated with the Home Ministry’s Crime and Criminal Tracking Network and Systems, allowing police to match FIRs of missing children against information stored in the Mission Vatsalya database.
Such integration helps create a search that works in two directions. In one State, the police may have the photograph attached to a missing-child FIR. Somewhere else, a Child Welfare Committee or Child Care Institution may have information about a child who has already been found. The problem is not about collecting information, but establishing whether two apparently unrelated records actually belong to the same child. Facial recognition technology has the potential to reduce that distance.
The basic principle of facial recognition in such cases is simple. Software analyses the face in one image and searches another collection of photographs to find visually similar faces. Instead of expecting a police officer to manually examine thousands of photographs, the technology can narrow the search to a much smaller group of possible matches.
While this reduces the workload on investigating agencies, a possible match cannot be treated as an identification. It is the responsibility of the police to establish whether the two photographs actually belong to the same person. Other details in the missing-person case, information from the child, family identification and the formal child-welfare process remain relevant before a search can turn into a reunion.
Today, Artificial Intelligence has brought tools that can match faces quickly and effectively. However, the use of facial recognition for missing children in India predates the current wave of AI tools.
On 5th September 2017, while hearing Bachpan Bachao Andolan v Union of India, the Supreme Court recorded that IT professional Vijay Gnanadesikan had developed a mobile application involving facial recognition for missing children. The petitioner told the Court that it would work with him so that the Government of India could use his expertise in developing and implementing such an application.
When the matter returned on 7th December 2017, the Supreme Court recorded that facial-recognition software was being actively considered by the Ministry of Women and Child Development. The Court left it to the Ministry to decide whether the software should be accepted and implemented.
The experiments that followed would produce one of the most widely repeated numbers in the Indian debate over facial recognition and missing children. They would also demonstrate why an algorithmic match and a child being reunited with the family cannot be treated as the same event.
In May 2018, the National Human Rights Commission referred to an exercise in which the Delhi government had run facial-recognition software against records of missing and found children. According to the NHRC, 2,930 missing-child records were matched with records in the found database on 20th April 2018. The Commission also highlighted a wider database problem, which was that police, Child Welfare Committees and Child Care Institutions sometimes entered names and places differently, making conventional matching difficult.
While the numbers were remarkable, what happened after that shows why they need careful description. On 22nd January 2019, the Delhi High Court recorded that it had been told the Facial Recognition Software being used by Delhi Police had not helped crack a missing-child case till then. The Court questioned why a system adopted after due diligence had apparently not produced the expected result.
A month later, Delhi Police told the High Court that the Facial Recognition Software was working efficiently. The Court, however, recorded concerns over the analysis of missing-child data placed before it and continued examining how the technology was actually translating into traced children.
By 26th April 2019, the Delhi High Court had a much more detailed breakdown before it. The Ministry of Women and Child Development had data relating to 11,093 children. After removing duplication, there were 10,617 probable matches. Of these, 3,202 children had been verified across India and were stated to be living in different Child Care Institutions. However, the Court recorded that only 127 pictures had actually been matched, meaning that the parents of 127 children had been traced. Even those children had not yet been restored at that stage. Another 6,415 children’s faces still had to be matched and verified.
This is why just repeating the line that facial recognition “found 2,930 missing children” is not enough. Once software matches the photographs and verification is done, police have to locate the original case and identify the family. Once all details are verified by the local police, the family has to be contacted to identify the child. Even if the identification is made, the child-welfare system needs to carry out due diligence before the child can be reunited with the family, including examining the circumstances of the disappearance and whether the child would be safe with the family.
A December 2023 Delhi High Court judgment recorded that the Face Recognition Application Software was functioning in the Crime Branch of Delhi Police and had been integrated with the online ZIPNET Missing Children Module. When the photograph of a missing child is uploaded, the software automatically searches recovered or found-child photographs, including data being uploaded nationally through the Ministry of Women and Child Development’s child-tracing system.
The search also works in reverse. When the photograph of a recovered or found child is uploaded, the system searches the missing-child photographs. The procedure recorded by the Court states that missing and found children’s photographs should be uploaded at the earliest opportunity and that investigators should obtain the probable results produced by the system in an attempt to link the two sets of records.
A police officer in one district does not have to remember a photograph circulated by another police station months earlier. The photograph itself can search for its counterpart.
Telangana Police’s Women Safety Wing maintains a technical analysis unit that assists in tracing missing and kidnapped people. Its official achievements page records several missing children traced with DARPAN, including a girl traced after two years, a boy traced after six years and other children whose cases had remained pending for years.
Telangana Police also operates a Missing Persons Monitoring Cell established with UNICEF, the Women and Child Development Department and NIC as stakeholders. The Women Safety Wing says the cell’s aim is to deal with missing-person cases using technology and to restore missing persons to their families or ensure rehabilitation where necessary.
The case of the boy who went missing in Handia shows why such a system is the need of the hour. Neither the disappearance nor the recovery happened in Telangana. The missing FIR was in Uttar Pradesh and the child was living in Assam. Telangana’s contribution was to run photographs through a system capable of suggesting that two records from those States might represent the same person.
In that case, technology did not rescue the boy. Assam Police had already done that five years before Telangana Police stepped in. Technology solved the information problem that remained after the rescue.
According to the Ministry of Home Affairs, the National Crime Records Bureau has launched UNIFY, an automated web-based photo-matching application using a machine-learning model. Police personnel can search photographs of missing persons, criminals and unidentified dead bodies against the national image repository stored in CCTNS.
CCTNS also has two national automatic alert services for missing-and-found person matches and missing-and-found vehicle matches. According to the Home Ministry, the alerts are generated and sent to the police stations concerned. Data from other national databases, including TrackChild-KhoyaPaya and the Railway Protection Force, are also available for search on the wider Inter-operable Criminal Justice System platform.
At the child-protection end, the Mission Vatsalya Portal is interoperable with CCTNS. The Women and Child Development Ministry says this allows missing-child FIRs to be matched with information in the Mission Vatsalya database by the police of States and Union Territories.
That means when these systems work together, the search no longer depends on a name, address or manually circulated photograph. A face can become a search query that can bridge gaps between investigation, recovery and reunion of the child.
However, there is an important gap in the system. According to the MHA, UNIFY does not provide a national child-specific figure showing how many searches have been conducted through the application, how many probable missing-child matches were produced, how many were confirmed and how many ultimately resulted in restoration. So, while India has a national automated photo-matching tool, it is much harder to measure how many missing children that particular tool has actually brought home.
There is an obvious difficulty when facial-recognition technology is being used for children. Children grow quickly and facial markers often change with time. A photograph submitted when a child is two may bear only limited resemblance to the same person a decade later. Hair, weight, facial proportions and other features change. Lighting, camera angle and the quality of the original photograph create further problems.
This is another area where artificial intelligence can assist investigators. When a child has remained missing for several years, the photograph available with the family or police may no longer resemble the person investigators are trying to find. AI-assisted age progression can use older photographs to generate an estimated image of how the child may look years later, giving police a more current visual reference to circulate through police stations, Child Care Institutions, railway networks, public places and other search channels.
Facial recognition and age progression therefore address different parts of the same problem. Facial recognition tries to determine whether a person who has already been found could be the child shown in an older photograph, while age progression attempts to estimate what the missing child may look like today. Neither can establish identity or guarantee recovery on its own, but both can give investigators new leads when an old photograph has become less useful because of the passage of time.
Railway stations already occupy an important place in India’s child-rescue system because a child can move from one State to another before investigators establish where to look.
The technological layer covering railway stations has expanded considerably. On 7th April 2026, the Ministry of Railways said Video Surveillance Systems had been expanded to 1,874 railway stations. According to the Ministry, the system used AI-based video analytics for automated detection of events such as intrusion and loitering and included Facial Recognition Software for real-time identification and monitoring.
By 24th July 2026, CCTV surveillance had been provided at 3,215 railway stations and 13,409 passenger coaches, according to another Railway Ministry reply. The Ministry said cameras at railway stations used AI-based video analytics capable of detecting events including intrusion, loitering and a fallen person.
However, that does not mean AI facial recognition was working at all 3,215 stations. The April 2026 statement explicitly said the Video Surveillance Systems expanded across 1,874 stations included Facial Recognition Software. The July statement describes the larger CCTV footprint of 3,215 stations and AI video analytics, but does not say that every one of those stations had facial-recognition capability.
A separate programme under the Nirbhaya Fund has also deployed an AI-based Facial Recognition System integrated with video surveillance and command-and-control infrastructure at seven major railway stations. The government identified the project as part of women’s safety measures implemented through the Ministry of Railways.
Even without an automated facial match, CCTV can change a missing-child investigation. Footage can establish when a child entered a station, the clothes the child was wearing, whether someone was accompanying the child and which platform or exit was used. If investigators identify a train, the search can move along the railway route rather than continuing without a direction.
In this sense, facial recognition is an additional layer on top of a much older investigative method. The camera provides an image that is used by the investigator to continue the search. In some cases, the investigation by the police is assisted by an algorithm that can speed up the process.
A missing child is not automatically a trafficking victim. But when the circumstances suggest trafficking, a different investigative chain begins. The Railway Protection Force launched Operation AAHT, Action Against Human Trafficking, and created Anti-Human Trafficking Units across the railway network. A Railway Ministry response said RPF AHTUs were operational at more than 740 locations and were coordinating with agencies involved in preventing trafficking and rescuing potential victims.
There is an important jurisdictional distinction. Railways has said that policing and public order are State subjects and that Government Railway Police or local police are responsible for the prevention, registration and investigation of crimes on the Railways. RPF supplements those agencies but is not empowered to investigate human-trafficking cases. When RPF detects suspected trafficking, the rescued persons and accused are handed over to the GRP or district police for further legal action.
Technology enters the trafficking response in ways that go beyond facial recognition. The Railway Ministry said RPF cyber cells had been instructed to patrol the web and social media for digital footprints and traces of human trafficking and to extract information useful for action against trafficking through the railway network.
RPF has also formalised cooperation with the Association for Voluntary Action, also known as Bachpan Bachao Andolan. The MoU signed in May 2022 provides for information sharing, capacity building, sensitisation and cooperation in identifying and detecting trafficking cases. That creates a very different technological chain from a facial-recognition search.
A trafficking operation can begin with intelligence, a digital footprint or suspicious movement. A train or passenger may then be identified, an interception organised and the rescued victim handed over to child-protection authorities while the police take over the criminal investigation.
On the morning of 13th May 2025, RPF acted on intelligence at Raxaul railway station in Bihar. A joint team of RPF, GRP Raxaul, the Sashastra Seema Bal Anti-Human Trafficking Unit, Railway Childline and Prayas Juvenile Aid Centre intercepted Train No. 15273, the Raxaul-Anand Vihar Satyagrah Express. Four girls aged between 13 and 17 were rescued. All four were from Nepal.
According to the Railway Ministry, the preliminary investigation indicated that the girls had been brought from Nepal with false promises involving employment and assistance in locating a missing relative in Gorakhpur. Their families had no knowledge of their journey. A suspected trafficker travelling with them was arrested, an FIR was registered at GRP Raxaul and the girls were handed over to child-protection authorities.
There was no facial-recognition breakthrough in the Raxaul operation. The technology trying to find or protect missing children is not one software package. Information has to reach the correct police or rescue unit at the correct time. In a trafficking case, the decisive factor may be an intelligence input rather than a face match.
The scale of railway anti-trafficking operations can be seen in RPF’s administrative data, although these figures should not be mixed with NCRB’s national missing-child statistics.
During 2024-25, Operation AAHT resulted in the rescue of 929 trafficking victims, including 874 children, according to the Railway Ministry. The 874 children included 824 boys and 50 girls. The same period saw 274 alleged traffickers arrested.
These Operation AAHT figures do not mean 874 children from the NCRB missing-child database were found through facial recognition, and they should not be used as evidence of facial recognition’s success. Different systems count different stages of the missing-child and trafficking response.
A recent Bengaluru case shows that a useful digital trace can come from an unexpected place. An 18-year-old woman named Tanishka disappeared from Bengaluru on 31st January 2026 after telling her family she was going to college with a friend. According to The Indian Express, she and her minor friend left behind their mobile phones and several personal belongings. Police searched more than 100 locations before the Karnataka High Court transferred the investigation to the Criminal Investigation Department.
Seven months later, CID investigators found a new lead when the woman used her Aadhaar details while applying for a government welfare scheme and entered a mobile number to receive an OTP. Investigators traced the number and eventually located her in a village in Karnataka.
The case should not be counted as a missing-child recovery because Tanishka was 18 when she disappeared. A person can avoid the digital trail police initially expect to find and still create another trace months later by interacting with a government service.
The experience with Delhi’s 2,930 matches makes more sense if facial recognition is understood as a series of stages. The first stage is the algorithmic stage, where the software says that two photographs may represent the same person.
The second stage is the investigation, where police check the missing FIR, age, physical details, location and other evidence and attempt to establish whether the suggested result is genuine.
Then comes the child-protection process. Even after a child’s identity and family are established, restoration may require further verification by the appropriate authorities.
The April 2019 Delhi High Court order shows exactly how different these stages can be. It recorded 10,617 probable matches, 3,202 children verified and 127 photographs where parents had actually been traced, while restoration still remained pending for those 127 at that point. A large number at the first stage therefore does not automatically become an equally large number at the final stage.
The machine reduces the number of faces investigators have to examine. Humans still have to establish what the match means.
The NHRC identified one of the fundamental problems with the older missing-child system in 2018. Information was being updated by police, Child Welfare Committees and Child Care Institutions, but minor variations in the spelling of children’s names or places could prevent records of a child listed as found from matching records of the same child listed as missing. The Commission cited facial recognition as a possible way to bridge that gap.
The current architecture attempts to connect more of these systems. Mission Vatsalya brings Child Welfare Committees, Juvenile Justice Boards, District Child Protection Units, Special Juvenile Police Units and Child Care Institutions onto an integrated digital platform, while the TrackChild component is connected with CCTNS.
On the police side, UNIFY searches the national CCTNS image repository and CCTNS can automatically send missing-and-found person match alerts to police stations.
Technology can help investigators, but it still depends on the quality of the data available. A system cannot search for a photograph that was never uploaded, fix every incomplete record or make sure that details of a child found in one district are immediately entered into the correct database. Facial recognition is only as useful as the information available to it.
There is also a question about what happens to a child’s photograph after the child is found. The Home Ministry says UNIFY allows police to search images of missing persons, criminals and unidentified dead bodies against the national CCTNS image repository. However, the publicly available information does not explain how long photographs or facial-matching data of recovered children are kept, or whether they are deleted or archived after restoration.
Facial recognition is also becoming part of a wider biometric system. On 19th June 2026, the Home Ministry launched NCRB’s CrPI application, which combines facial recognition, iris matching and DNA matching. Its stated uses include identifying repeat offenders and matching suspects seen in CCTV footage. At the time of launch, 2,190 CrPI units had been distributed and more than 5.53 lakh enrolments had been completed. The government has not said that CrPI is currently being used to find missing children, but it shows how police identification systems are becoming more advanced.
Delhi Police is expanding its technological systems as well. In March 2026, C-DOT signed an MoU with Delhi Police covering facial recognition, secure communications and other policing technologies. Less than a decade after the Supreme Court was discussing whether facial recognition for missing children should even be explored, such technology is now part of several police and surveillance systems.
The harder question is how much difference these systems actually make. India publishes figures for children reported missing and later traced. RPF publishes rescue and trafficking figures. Mission Vatsalya can connect missing and found child records, CCTNS can issue missing-and-found alerts, and UNIFY can search photographs against the national image repository.
What is not easily available is a national figure showing how many missing children were found specifically because of facial recognition, CCTV analytics or automated photo matching. Government material on UNIFY explains what the system can do, but does not publish child-specific figures for searches, confirmed matches or restorations. Railway and Mission Vatsalya data also do not clearly show how many recoveries happened because a particular technology supplied the key lead.
The boy from Handia shows both the strength and the limits of such technology. He was found in Assam just nine days after disappearing from Uttar Pradesh, but the connection between the two records was made only five years later through facial recognition. Telangana Police produced the possible match, Uttar Pradesh Police contacted the family and his parents travelled to Assam to identify him.
Technology did not replace the police, child-welfare authorities or the family. It simply connected information that had remained separated for years. That may be its most useful role in the search for missing children.
For decades, India’s economic potential was bottlenecked by a railway network that forced heavy freight and premium passenger expresses to share the same congested tracks. Today, a monumental paradigm shift is underway. Through the operationalisation of Dedicated Freight Corridors (DFC), the multitracking of high-density routes, and the deployment of 12,000 HP locomotives, Indian Railways is physically segregating its traffic. This deep dive explores the engineering and economic overhaul—from double-decker freight to the Kavach safety shield—that is fundamentally rewriting the nation’s supply chain geometry.
An exploration contract allows a country to study an area, map its geology and collect samples. It does not give permission to start mining. At present, no country has an approved commercial exploitation contract for minerals in the international seabed, including India.
NCRB data, CCTNS, TrackChild, Khoya-Paya, GHAR, helplines and Railway operations form a layered search system where finding a child is only the first step before verification, protection and safe restoration.
Selected issues, clear explanations and thoughtful perspectives—delivered when we publish.
