Xuemei Cheng

A novel integrated PCA and FLD method on hyperspectral image feature extraction for cucumber chilling damage inspection

High-resolution hyperspectral imaging (HSI) provides an abundance of spectral data for feature analysis in image processing. Usually, the amount of information contained in hyperspectral images is excessive and redundant, and data mining for waveband selection is needed. In applications such as fruit and vegetable defect inspections, effective spectral combination and data fusing methods are required in order to select a few optimal wavelengths without losing the crucial information in the original hyperspectral data. In this article, we present a novel method that combines principal component analysis (PCA) and Fisher’s linear discriminant (FLD) method to show that the hybrid PCA-FLD method maximizes the representation and classification effects on the extracted new feature bands. The method is applied to the detection of chilling injury on cucumbers. Based on tests on different types of samples, results show that this new integrated PCA-FLD method outperforms the PCA and FLD methods when they are used separately for classifications. This method adds a new tool for the multivariate analysis of hyperspectral images and can be extended to other hyperspectral imaging applications for fruit and vegetable safety and quality inspections.

NIR/MIR DUAL–SENSOR MACHINE VISION SYSTEM FOR ONLINE APPLE STEM–END/CALYX RECOGNITION

A near–infrared (NIR) and mid–infrared (MIR) dual–camera imaging approach for online apple stem–end/calyx detection is presented in this article. How to distinguish the stem–end/calyx from a true defect is a persistent problem in apple defect sorting systems. In a single–camera NIR approach, the stem–end/calyx of an apple is usually confused with true defects and is often mistakenly sorted. In order to solve this problem, a dual–camera NIR/MIR imaging method was developed. The MIR camera can identify only the stem–end/calyx parts of the fruit, while the NIR camera can identify both the stem–end/calyx portions and the true defects on the apple. A fast algorithm has been developed to process the NIR and MIR images. Online test results show that a 100% recognition rate for good apples and a 92% recognition rate for defective apples were achieved using this method. The dual–camera imaging system has great potential for reliable online sorting of apples for defects.

High-resolution real-time x-ray and 3D imaging for physical contamination detection in deboned poultry meat

This paper describes a novel approach for detection of foreign materials in deboned poultry patties based on real-time imaging technologies. Uneven thickness of poultry patties could lead to a significant classification error in a typical X-ray imaging system, and we addressed this issue successfully by fusing laser range imaging (3D imaging) into the x-ray inspection system. In order for this synergic technology to work effectively for on-line industrial applications, the vision system should be able to identify various physical contaminations automatically and have viable real-time capabilities. To meet these challenges, a rule-based approach was formulated under a unified framework for detection of diversified subjects, and a multithread scheme was developed for real-time image processing. Algorithms of data fusion, feature extraction and pattern classification of this approach are described in this paper. Detection performance and overall throughput of the system are also discussed.

Hyperspectral Imaging and Feature Extraction Methods in Fruit and Vegetable Defect Inspection

High-resolution hyperspectral imaging (HSI) provides an abundance of spectral data for feature analysis in image processing. Usually, the amount of information contained in hyperspectral images is excessive and redundant, and data mining in waveband selection is needed. In applications for fruit and vegetable damage inspection, effective spectral combination and data fusing methods are required in order to select a few optimal wavelengths without losing key information in the original HSI data. In this paper, we present a new method that combines the principal component analysis (PCA) and fisher’s linear discriminant (FLD) method in a way that maximizes the representation and classification effects on the extracted new feature bands. The method is applied to the detection of chilling injury on cucumbers. Compared with PCA and FLD when used separately, this new integrated PCA-FLD method has achieved results showing better classification performance when tested on different types of samples. This method is ready to be extended to other hyperspectral imaging applications for fruit and vegetable safety and quality inspections.

Integrated PCA-FLD method for hyperspectral imagery feature extraction and band selection

An important task in hyperspectral data processing is to reduce the redundancy of the spectral and spatial information without losing any valuable details that are needed for the subsequent detection, discrimination and classification processes. Band selection and combination not only serves as the first step of hyperspectral data processing that leads to a significant reduction of computational complexity, but also a invaluable research tool to identify optimal spectra for different online applications. In this paper, an integrated PCA and Fisher linear discriminant (FLD) method is proposed for hyperspectral feature band selection and combination. Based on tests in a hyperspectral detection application, this new method achieves better performance than other feature extraction and selection methods in terms of robust classification

Real-time image analysis for nondestructive detection of metal sliver in packed food

Foreign materials such as metal slivers and stones in packed food are listed safety hazards, which could lead to severe health problems. In this paper, a real time X-ray imaging inspection method is investigated for foreign material detection in chili packages. A new image segmentation method combining edge detection and region growing was successfully applied to address the challenges due to the uneven thickness of chili package.

Real-Time Image Analysis for Nondestructive Detection of Metal Slivers in Packed Food

Foreign materials such as metal slivers and stones in packed food are known safety hazards that can lead to severe health problems. In this article, a real-time x-ray imaging inspection method is investigated for foreign material detection in chili mix packages. A novel image segmentation method combining Canny edge detection and region growing was successfully applied to overcome the problem caused by uneven thickness of chili mix packages and discriminate false positives caused by bubbles and foil wrinkles. The detection and false positive rates for metal slivers as small as 0.5 × 0.5 × 1 mm were 98.1% and 0.64%, respectively.

Gabor-wavelet decomposition and integrated PCA-FLD method for texture based defect classification

In many hyperspectral applications, it is desirable to extract the texture features for pattern classification. Texture refers to replications, symmetry of certain patterns. In a set of hyperspectral images, the differences of image textures often imply changes in the physical and chemical properties on or underneath the surface. In this paper, we utilize Gabor wavelet based texture analysis method for textural pattern extraction, and combined with integrated PCA-FLD method for hyperspectral band selection in the application of classifying chilling damaged cucumbers from normal ones. The classification performances are compared and analyzed.

Feature extraction and band selection methods for hyperspectral imagery applied for identifying defects

An important task in hyperspectral data processing is to reduce the redundancy of the spectral and spatial information without losing any valuable details that are needed for the subsequent detection, discrimination and classification processes. Band selection and combination not only serves as the first step of hyperspectral data processing that leads to a significant decrease in computational complexity in the successive procedures, but also a research tool for determining optimal spectra requirements for different online applications. In order to uniquely characterize the materials of interest, band selection criteria for optimal band was defined. An integrated PCA and Fisher linear discriminant (FLD) method has been developed based on the criteria that used for hyperspectral feature band selection and combination. This method has been compared with other feature extraction and selection methods when applied to detect apple defects, and the performance of each method was evaluated and compared based on the detection results.

Pattern classification for boneless poultry inspection using combined X-ray/laser 3D imaging

A combined X-ray/laser 3D imaging technology has been developed for bone fragment and foreign material detection in boneless poultry products. In this paper, various methods of pattern classification including neural network and statistical approaches are applied to the poultry images obtained by the combined imaging system, and the classification performances are compared and analyzed.