·       Neural Network Inspection Systems

EnVision 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.

 

EnVision 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 re-engineering.

 

EnVision 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.

 

A 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.

 

As 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.

 

EnVision Systems LLC is available for consulting, software contracting, and system development for OEM and end user; for image analysis and Windows user interface.