· Neural Network Inspection Systems
Systems LLC is dedicated to bringing cutting edge image processing and pattern
recognition technology to industrial image inspection at a low cost.
We pride ourselves on our extensive experience with state of the art
image analysis and pattern recognition. We
deliver this technology at a low cost by using a standard Microsoft Windows PC
software platform rather than expensive proprietary image processing and
pattern recognition hardware. Our
software is implemented efficiently to eliminate the need for expensive
proprietary image processing hardware.
Systems LLC specializes in the use and creation of image processing
neural network pattern recognition technology. Neural
networks is the most powerful pattern recognition tool available today;
combining great accuracy, speed, robustness, and flexibility.
Neural networks free the engineer from the impossible and time
consuming task of trying to deduce and codify the exact relationship between
the data and the observed results. If
the relationship changes, the neural network can be easily re-trained without
incurring the high recurrent engineering costs associated with rule base
Systems LLC has employed custom neural network systems in product sorting,
Q&A, aerospace, medical image processing and many other industrial image
inspection applications using diverse criteria including texture,
multi-spectral data, morphology, subjective criteria, etc.
typical neural network learns to do pattern classification by being presented
with product examples of each class. Explicit
knowledge of how features of a product relate to its class is unnecessary.
Simply gather a representative sample of the product.
Train the neural network by presenting it data extracted from these
samples with an indication of each sample’s class. The network will learn how to determine the class by itself.
The network is then capable of classifying previously unseen product.
an example, suppose that the application involves grading a type of fruit
using three band color images and a neural network system has been trained to
recognize green and black spots. If
a fruit sample has either of these spot colors over a certain size then it is
desired to assign them a particular grade.
As examples of fruit are being collected, it is determined that fruit
which has significant amounts of both green and black spot should also be
assigned that grade regardless of whether or not either spot color achieves
the size threshold. It is
additionally determined that spot shape should also effect the grade.
The situation becomes even more complicated as more spot types are
added. It is virtually impossible
to quantify and so program this complex relationship.
By simply presenting a neural network with training data obtained from
these examples (with the desired grades), the neural network will learn how to
assign the proper grades.
Systems LLC is available for consulting, software contracting, and system
development for OEM and end user; for image analysis and Windows user