Can the UK Predict Crimes on Buses?
Though many are criticizing a UK research project, a deeper look shows that the project is technically feasible though faces privacy concerns. For background, read the BBC's article.
The concerns are (1) this cannot technically be done and (2) this risks falsely accusing individuals. Adding to this is the use of the word 'predict' which creates silly and unrealistic expectations. In reality, these systems identify potentially suspicious behavior that improves the efficiency of human operators' detecting actual crimes.
The system will combine a number of sensors (including video and audio) as well as GPS information and historical information about crime rates in areas (see an article with details [link no longer available] and a whitepaper from the researchers).
While the system may use video analytics to detect gender, personal characteristics and types of behavior, this is only one of 3 key components. Additionally, the system will cross-reference the video/audio elements with statistical analysis of how historically dangerous the given time and place of the activity is (i.e., 3 am in a high crime neighborhood). Next, the system plans to send notices to the bus drivers and/or central command to review 'suspicious' activities.
Finally, the system is not scheduled for production deployment for 5 years - 2014, providing ample time for maturity and cost reduction of the various analytics to be used.
Unlike perimeter detection systems common today, this system's alerts would not need to be 100% accurate as they would be designed for operator review similar to how manual operators today scan video.
The largest concern may be the use of profiling people based on gender, race, physical characteristics, etc. and locations based on crime levels (though the articles do not explicitly mention race/ethnicity, clearly this will be an issue addressed). Moreover, given the UK's passionate debate about CCTV and privacy this could become a central obstacle in the future use of this system.