Why ALPR Technology Accuracy Matters
When considering ALPR solutions, accuracy is a key–if not the top–consideration, and ALPR+ is one of the most reliable solutions on the market. In a previous blog, we dug into what the “plus” is all about; simply put, MMCG (make, model, color, generation) collects more information than standard ALPR solutions. But what makes the technology so accurate is “Deep Learning”-based technology.
Deep Learning is a branch of Artificial Intelligence (AI) that focuses on training machines using substantial computer power (GPUs) with large amounts of labeled data. It also removes the manual step of feature selection and provides the system full autonomy on how the model should be trained. Traditional methods simply cannot compete with the advancements and achievements of Deep Learning solutions present in ALPR+.
Precision, Recall, and Flexibility: Finding the right balance
When considering ALPR solutions, there is always a tradeoff between precision and recall. Precision takes into account the amount of false positives generated while recall is a measure of how many actual objects we missed.
Both these metrics are adjustable and dependent on the confidence threshold we set for accepting or rejecting results. Lower threshold values would result in more false positives (low precision) but make sure no objects are missed (high recall). As the confidence threshold increases, you start to reduce the false positives (high precision) at the expense of more missed objects (low recall). Achieving the right balance of precision and recall can be the difference between a good system and great system.
Another key component to consider is solution flexibility. End users need the power to adjust their ALPR solution to specific use cases to ensure it works for their organization’s specific needs.
For ALPR providers, rigorous, thorough testing over multiple types of challenging environments helps pinpoint areas to improve. And speaking from Sighthound’s personal experience, all this testing leads to a much more accurate, effective product–enter ALPR+. Read on to see how a company uses ALPR+ to boost productivity, profits, and make their customers happy.
Case Study: Going from 15% to 90% Vehicle Identification Accuracy
We’ve established that accuracy is crucial in an LPR system, as is precision, recall, and flexibility. But what does all of this look like in action? Implementing a system with capabilities that goes beyond standard (and often inaccurate) license plate recognition software can mean a big boost to your bottom line.
Take a global provider of automated car alignment solutions made huge changes to their bottom line. They were using a standard LPR system to extract details of cars, including the make, model, and color, to ensure they properly aligned every vehicle. The problem with this process was that additions or modifications in license plate formats and designs weren’t recognized by their LPR software, and this resulted in a significant drop in LPR accuracy. Because of this, more manual input was required from the service staff, taking up valuable time and resources.
For example, this company worked with customers in Tennessee, where license plate designs were recently modified. Because of the two license plate formats, he existing LPR system was not effectively collecting all the information their customers required–even going down to just 15% accuracy for part of the business.
To deliver the highest quality of services, they needed a solution that would be able to identify vehicles without relying on license plate consistency. They implemented Sighthound ALPR+ to ensure their performance, their customers, and their profit didn’t suffer as they were able to recognize cars with greater than 90% accuracy–without relying on their legacy LPR system.
Keeping Profits Rolling with ALPR+ and Robust MMCG Data
With the data collected from Sighthound ALPR+, the company has automated mechanical adjustments for car service equipment based on license plate reading and collecting vehicle make, model, color, and generation. They’re saving company resources and valuable employee time by automating data collection.
Now, they have confidence that with ALPR+, they are collecting vehicle information with over 90% accuracy.
In the three years of using ALPR+, they have:
Expanded the solution from the EU to North America, including the U.S. and Canada
Maintained over 90% accuracy in vehicle identification
Have received exceptional reviews from their customers on how robust the system is at recognizing cars in different scenarios (lighting, angles, distance/size, image quality, car modifications, etc)
Want to learn more?
Sighthound ALPR+ users are constantly finding innovative ways to utilize insights from data–especially MMCG–and create more value in their business. Want to learn what ALPR+ could mean for your business?Schedule a demo or request a free trial to see how ALPR+ can make your organization more agile and improve your bottom line.