Location: Sugarbeet and Bean Research2013 Annual Report
1a. Objectives (from AD-416):
This research will develop practical sensors and technologies for quality measurement and grading of fruits and vegetables before, at and after harvest. It also aims to generate new knowledge and understanding of the optical and mechanical properties of fruits and vegetables and their relationship with the physiological factors and quality attributes. A systems approach of integrating sensors development, properties characterization, and models/algorithms development will be applied to attain the following specific objectives: Objective 1: Develop cost effective sensors and sensing systems to measure and monitor the quality/maturity of individual apples in the orchard. Objective 2: Develop commercially viable technology to presort and grade apples in the orchard so as to decrease postharvest handling and storage costs for fruit growers. Objective 3: Develop technology to accurately and rapidly assess, sort, and grade harvested tree fruits and vegetables for multiple internal quality attributes (firmness, flavor, ripeness) and defects.
1b. Approach (from AD-416):
First, latest miniature near-infrared (NIR) technology will be used in the development of a portable fruit maturity sensor. Fruit maturity measurement will be achieved through integration of NIR technology with the nondestructive firmness measurement method recently developed in our lab. Algorithms will be developed and integrated into the sensor for real-time measurement of fruit firmness, soluble solids content and other maturity parameters. Laboratory and field tests will be performed to assess the sensor’s performance and portability. Second, commercially viable infield mobile sorting technology will be developed for sorting and grading harvested apples into two or three quality grades (fresh market, processing, and cull). A cost effective machine vision system and a fruit bin filler will be designed and built; they will then be integrated into existing infield apple handling systems for effective segregation of unmarketable or defective fruit from fresh market grade fruit. Laboratory and infield tests will be performed to evaluate the mobile infield sorting prototype for performance and bruising to apples. We will collaborate with a horticultural equipment manufacturer and university extension specialists, so that the developed technology can be quickly adopted by growers to achieve production cost savings. Third, research will be conducted on the development of a hyperspectral imaging-based spatially-resolved method for measuring the optical absorption and scattering properties of horticultural and food products. An optical property measuring prototype will be built and tested for automatic measurement of the optical properties of fruits and other food and agricultural products. Optimization of the hardware (light source, source-detector distance, etc.) and algorithms will be performed through Monte Carlo simulation and experiment to improve measurement accuracy and repeatability. Experiments and mathematical or statistical analyses will be performed to relate the optical properties to the structural/mechanical properties of apples and to the quality attributes of apples, peaches and tomatoes. Moreover, research will be conducted to improve spectral scattering technology for sorting and grading apples for firmness and soluble solids content. Improved hardware designs and new spectral scattering analysis methods integrating both spectral and image features will be considered and incorporated into the laboratory spectral scattering prototype for classification of apples into different quality grades based on firmness/soluble solids content. Finally, an online hyperspectral imaging system, which integrates reflectance in the visible spectral region and transmittance in the short-wave near-infrared region, will be developed for online sorting and grading of pickling cucumbers and/or pickled products for external and internal quality (or defect). Different lighting designs and spectral imaging acquisition modalities will be considered and evaluated. Image processing and analysis algorithms will be developed for rapid detection and segregation of defective pickling cucumbers/pickles.
3. Progress Report:
Spectral scattering is useful for nondestructive sensing of fruit firmness. The technique, however, depends on appropriate quantification of scattering features and the development of a reliable statistical prediction model relating the scattering features to fruit firmness. A new method, called moment method, was proposed to extract important features from the spectral scattering images, and it was evaluated for ‘Delicious’, ‘Golden Delicious’ and ‘Jonagold’ apples. The method resulted in consistently better prediction of fruit firmness, compared with the mean reflectance method used in previous studies. Since spectral scattering for firmness prediction is influenced by such factors as the variability of firmness in the calibration samples, data processing method and harvest season, research was carried out to evaluate the effect of these factors and their interactions on the performance of the firmness prediction models for three cultivars of apple. The model performance generally improved with the increasing number of samples used for building the prediction model. Overall about 400 samples with a representative range of firmness were needed for building a robust firmness prediction model. Progress was made on the development of a mobile system for harvesting and sorting apples in the orchard. Harvest aid functions that are suitable for six to eight workers were incorporated into the mobile system, which enhance harvest efficiency and improve worker safety. Improvements to the bin filler designs were made for better delivery and distribution of harvested fruit into the bins. New functions and algorithms were developed and incorporated into the apple sorting/grading software program to provide more user-friendly interfacing in selecting grading criteria and quality grades. Research was carried out to improve the hyperspectral imaging system, operated in simultaneous reflectance and transmittance modes, for online inspection of both external and internal quality of pickling cucumbers. Two optimal wavelengths or wavebands were determined using two different wavelengths selection methods (i.e., minimum redundancy-maximum relevance and principal component analysis). Superior results in segregating defective cucumbers from normal ones were obtained, with an accuracy of greater than 94%. The identified wavebands can be implemented for fast online inspection of internal defect of pickling cucumbers. Firmness and soluble solids content (SSC) are important in assessing the quality and shelf life of blueberries. Hyperspectral reflectance and transmittance images were acquired from blueberries over the wavelengths of 500-1,000 nm. Statistical models were developed for prediction of firmness and SSC. Better predictions of SSC and firmness were obtained using reflectance mode than transmittance mode. Fruit orientation only had small effect on reflectance measurement and insignificant effect on transmittance measurement.
1. Optimization of spectral scattering measurement for fruit quality assessment. Spectral scattering technique provides an effective means for measuring light scattering in food products like apple, and scattering features are useful for predicting firmness and soluble solid content in fruit, two quality attributes that are important to the consumer. Appropriate quantification of scattering features and development of a proper statistical model are critical to accurate prediction of fruit firmness and other quality attributes. A new method, called moment method, was proposed and evaluated by ARS researchers in East Lansing, Michigan for describing spectral scattering features of apples. The method performed consistently better in predicting the firmness of three cultivars of apple, compared with the mean reflectance method used in previous studies. Since spectral scattering prediction of apple firmness is also influenced by the variability of firmness in the calibration samples, data processing method and harvest season, optimization of these factors was performed, which resulted in important recommendations on appropriate selection of these factors so as to achieve superior spectral scattering prediction of fruit firmness. These recommendations will help researchers and equipment developers better implement spectral scattering technique for quality assessment of fruit.
2. Spectral imaging classification of normal and defective pickling cucumbers. Pickling cucumbers are susceptible to internal damage due to adverse climate and growth conditions, pest and/or disease infestation, and improper harvest and postharvest operations. Effective detection and removal of defective cucumbers prior to brining is critical to the quality of final pickled products. ARS researchers at East Lansing, Michigan developed a hyperspectral imaging technique for detecting both external and internal quality defects of fresh pickling cucumbers. The technique is, however, not fast enough for online inspection of pickling cucumbers because it needs to acquire and process a large quantity of spectral image data. Research was thus carried out to improve the lighting configuration for better transmittance measurement of pickling cucumbers and to determine the optimal wavelengths or wavebands to meet the high speed inspection requirement. Hyperspectral reflectance and transmittance images of pickling cucumbers were acquired and analyzed to select the optimal wavebands for detecting defective cucumbers from normal ones. It was found that using the ratio of two optimal wavebands in the near-infrared region achieved more than 94% classification accuracy in differentiating normal and defective cucumbers. The identified wavebands can be implemented for rapid, real-time detection of defective pickling cucumbers, which will provide the pickle processor an effective inspection tool to ensure quality of pickled products.
Huang, M., Wang, B., Zhu, Q., Lu, R. 2012. Analysis of hyperspectral scattering images using locally linear embedding algorithm for apple mealiness classification. Computers and Electronics in Agriculture. 89:175-181.