Image analysis – Applications – Animal – plant – or food inspection
Reexamination Certificate
1997-09-02
2001-04-17
Mancuso, Joseph (Department: 2723)
Image analysis
Applications
Animal, plant, or food inspection
Reexamination Certificate
active
06219438
ABSTRACT:
FIELD OF THE INVENTION
The invention relates to a system and method of identifying objects, and, more particularly, a system and method of identifying fruits and vegetables based on their respective unique surface textures sensed by an optical scanner.
BACKGROUND OF THE INVENTION
Modern supermarkets rely on relatively fast checkout procedures to maintain competitive advantages over relatively smaller scale grocers. Checkout usually involves determining a customer's total bill and receiving payment for purchased products corresponding to the bill. Many factors can influence improvements in customer throughput, which are directed to the overall goal of maximizing sales and reducing the time a consumer spends waiting in line.
By maximizing the volume of products sold, grocery chains tend to compete more effectively because overall prices may be lowered to attract a greater number of shoppers and still maintain acceptable profits. Moreover, in reducing the time a consumer waits in line at the checkout stand, customer satisfaction increases, tending to improve the probability of that customer returning. One of the keys to the grocery industry's success in maximizing customer throughput and minimizing checkout times involves a coded label known as a bar code.
A typical bar code includes a two-dimensional pattern comprising horizontally adjacent black and white vertical bars of varying widths. The respective widths and positions correspond to a coded identification number representing a particular item. The code is printed on a label that is affixed to the item, eliminating the need for a separate price tag. At the time of purchase, instead of manually inputting a price using register keys, the grocery clerk merely positions the bar code within the field of view of a bar code scanner. The scanner reads the code, automatically rings up the product price, and causes the name of the item and the price to print out on a convenient receipt.
Conventional bar code readers generally include a laser beam directed along a scanning path by a rotating mirror. A photodetector, positioned within the scanner and oriented substantially along the path of light reflected from the bar code and received within the scanner beam, detects the reflections of the beam from a targeted surface. When the scanner targets a bar code, the reflected components of the beam vary in intensity depending on whether the reflection was incident a white or black bar. An input filter analyzes the data to determine components that exceed respective positive and negative thresholds corresponding to transitions between black and white bars. The coded transitions are then decoded by a bar code processor to command the register to automatically ring up the product and its corresponding price.
While conventional bar code scanners are beneficial for their intended purposes, i.e. to dramatically improve supermarket checkout times, such scanners typically only identify the contents of bar codes. This proves problematic for produce. As a result, most checkout stand equipment typically requires the operator or clerk to manually input a code identifying the many fruits and vegetables purchased and the corresponding price. Unfortunately, this procedure requires a substantial amount of training for the checkout stand clerks to properly recognize and identify the fruits and vegetables. Even with the investment in costly training, error rates often run as high as twenty percent. Moreover, the procedure frequently involves significant delays, since the clerk often must refer to a price table to match the items to the proper prices.
A limited solution to the above problem involves affixing individual labels to all fruits and vegetables. These labels typically comprise a human-readable name or number, and may include a bar code. Such labels tend to eliminate most identification errors, but fail to speed up the checkout procedure.
Although applying labels to individual articles of produce appears beneficial in some instances, problems exist for many fruits and vegetables. In one aspect, the sheer numbers of fruits and vegetables create significant labor costs in order to effect the labeling procedure. Much of the cost arises from assigning employees to individually apply the labels to sizeable quantities of produce. Further, label removal by customers from fruits and vegetables having delicate surfaces often damages those surfaces due to the adhesive associated with the labels.
Because of the problems described above, those skilled in the art have recognized the need for a system and method of identifying produce in a timely and efficient manner in order to minimize checkout times, maximize pricing accuracy, and eliminate unnecessary food waste. The system and method of our present invention satisfies these needs.
SUMMARY OF THE INVENTION
The system and method of our present invention provide the capability of identifying produce without the need for extraneous labels or other attachable identification indicia. By analyzing the unique surface texture of the produce itself, the invention substantially reduces costs by minimizing labor and pricing errors during checkout. Additionally, checkout times are measurably improved through the elimination of unnecessary delays attributable to finding and properly positioning a label for reading by a clerk or scanning. Consumer satisfaction will also increase through the elimination of the need to remove potentially damaging adhesive labels.
To realize the advantages summarized above, the system of the present invention identifies objects having respective physical surface properties, such as texture, and includes a computer having a memory and an optical scanner responsive to the computer. The scanner has an optical source for generating a laser beam of a predetermined wavelength for scanning over the textured surface and producing reflected intensity data representative of the textured surface. An optical signal detector senses the reflections of the beam incident upon the surface. Means are provided for transforming the intensity data so that a feature vector, comprising a plurality of vector components, can be extracted. The components have respective coefficients corresponding to at least one physical property representative of the surface, e.g. its texture. Also provided in the system is a means for comparing the extracted feature vector to a set of predetermined feature vectors stored in the memory and representing a plurality of object types in order to determine whether the article has a surface feature similar to the surface of any one of the objects represented by the predetermined vectors included in the set.
The method of the present invention identifies an object having respective surface properties, e.g. texture, with an optical scanner. The scanner has a laser beam and an optical signal detector coupled to a computer. More particularly, the method includes the steps of first acquiring data representative of the intensity of reflections of the beam incident on the textured surface of the object. The method proceeds by transforming the intensity data so as to extract a translation invariant physical feature vector comprising a plurality of vector components. The vector components include respective coefficients corresponding to at least one physical property representative of the surface. The method then concludes by comparing the extracted feature vector to a set of predetermined feature vectors representing a plurality of object types to determine whether the article has a surface similar to any one of the objects represented by the predetermined vectors included in the set.
REFERENCES:
patent: 4737857 (1988-04-01), Rucci et al.
patent: 4881068 (1989-11-01), Korevaar et al.
patent: 4881818 (1989-11-01), Bustamante et al.
patent: 4988852 (1991-01-01), Krishman
patent: 5214470 (1993-05-01), Denber
patent: 5220617 (1993-06-01), Bird et al.
patent: 5325301 (1994-06-01), Knoff et al.
patent: 5392255 (1995-02-01), LeBras et al.
patent: 5426506 (1995-06-01), Ellingson et al.
patent:
Giordano David A.
Kochanski Gregory P.
Cooperrider F. E.
Lucent Technologies - Inc.
Mancuso Joseph
LandOfFree
Produce indentifier using barcode scanner and wavelet image... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Produce indentifier using barcode scanner and wavelet image..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Produce indentifier using barcode scanner and wavelet image... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2457292