Ahead of the Curve: Machine Learning Transforms Parking Guidance Technology
In science fiction novels, artificial intelligence (AI) conjures images of out-of-control computers destroying humanity and taking over the earth. It’s scary stuff. Fortunately, the reality is much less terrifying—and much more beneficial to parkers and parking owners and operators. One of the most significant recent breakthroughs in parking guidance technology is machine learning. Through machine learning, parking guidance has become more accurate and more useful for both parking inventory management and curb management.
Dynamic Learning Capability
In 1959, Arthur Samuel, a pioneer in the AI field called machine learning “a field of study that gives computers the ability to learn without being explicitly programmed.” Machine learning is a type of AI that allows equipment to modify itself when exposed to more data. Because the process is dynamic and doesn’t require human programmers or designers to make certain changes, machine learning elements continually improve the equipment’s utility. The process of machine learning can be compared to the cognitive process of a baby. It is born knowing nothing and adjusts its understanding of the world in response to experience.
Algorithms for Accuracy
Machine learning uses algorithms to predict outcomes and to enhance the accuracy of those predictions. The technology is programmed into parking guidance equipment not only to process the video frames but also to make decisions based on those frames. This is done with a neural network. With neural networks, systems continually measure errors and modify the programs in order to minimize—and hopefully eliminate—those errors. Neural networks allow technology to learn a task by analyzing training examples. An object recognition system might be fed thousands of labeled images of cars, trucks, motorcycles, buses so it can find visual patterns that consistently correlate with particular labels. A system trained to recognize vehicles can search an image for all instances of certain items and then provide the number and location of those items in the image.
Machine Learning and Parking
This type of machine learning approach typically demands extra and specialized resources. Because of that, it has only started to become a possibility with parking guidance systems (PGS) in the past few years. Yet, it is a much more accurate and precise approach for obtaining detailed information about what is in an image. This is important so that facility access, parking location, and vehicle volume can be correctly tracked. For example, a system with machine learning that is trained to recognize vehicles will first take and image and look for all instances of the items. Then it will let the system know how many of those objects it found and where they are in the image. When it comes to identifying different vehicles, the neural network learns the dominant identifying features of different kinds of vehicles, including cars, motorcycles, scooters, and delivery trucks.
System Quality Assurance
Machine learning can also provide a confidence score associated with the objects to ensure that neural networks learn the correct information. If an object is detected, but has a low score, it is still detected. While the guidance may not include it as an object of interest initially, in can later evaluate all the objects that had a low score to see why the score was low. he advantage of this is that it allows the system to recover from low scores and continually train the network to get smarter over time. It’s an important feature that allows the system to continually improve itself.
The more accurate a PGS is at reading the types of vehicles in the facility, the better able the system is to offer accurate occupancy information. The benchmark for single space guidance (SSG) has traditionally been 99%. SSG providers have relied on this level of accuracy to persuade owners and institutions to spend tens of thousands—sometimes hundreds of thousands—of dollars for SSG. With the emergence of machine learning, intelligent cameras, that provide zone and level counting have reshaped the accuracy that can be expected and achieved with these systems.
Practical Considerations for AI
While the goal has always been 99% per counting node, few systems actually achieved that. Now, with some purpose-built, vision-based technologies, each node can consistently achieve 90% or more through continuous improvement with machine learning. This is particularly important when lots and garages are expected to provide service in addition to parking vehicles. For instance, many facilities offer space for delivery vehicles. As curb management is key to mobility, a PGS’s ability to recognize different vehicles is essential to quantify occupancy and space availability. This enables better curb management in an automated way.
Accuracy Cost Benefits
Improved accuracy through machine learning also offers significant cost benefits to parking owners and operators. In the past, owners and operators had to install SSG technology to achieve sufficient accuracy. The challenge was that cost per space could run between $250 and $750 for a single space system. That meant that for a garage with just 500 spaces, installation of a reliable guidance system could cost $125,000 to $375,000. And that’s just the installation. There may also be maintenance and repair costs to factor in. With machine learning, significant improvements in parking guidance accuracy can be made without the need to install expensive and intrusive infrastructure. This is particularly important for owners who are considering retrofitting parking facilities with PGS. In fact, the technology can often reuse existing camera infrastructure and guidance signage that may have been installed with an antiquated count system. Simple systems that rely on intelligent cameras with machine learning and main entrances and exits, individual floors, and sections of floors provide accuracy once available only with SSG. And it can be done for a fraction of the cost.
Future of Parking Guidance
Machine learning represents the future of parking guidance. The AI elements of machine learning enable parking guidance to be more efficient and cost-effective and to play a more central role in promoting mobility through curb management.