Peter P. Ling

Visual feedback guided robotic cherry tomato harvesting

Harvesting cherry tomatoes is more laborious than harvesting larger size tomatoes because of the high fruit density in every cluster. To save labor costs, robotic harvesting of cherry tomatoes has been studied in Japan. An effective vision algorithm, to detect positions of many small fruits, was developed for guidance of robotically harvested cherry tomatoes. A spectral reflectance in the visible region was identified and extracted to provide high contrast images for the fruit cluster identification. The 3-D position of each fruit cluster was determined using a binocular stereo vision technique. The robot harvested one fruit at a time and the position of the next target fruit was updated based on a newly acquired image and the latest manipulator position. The experimental results showed that this visual feedback control based harvesting method was effective, with a success rate of 70%.

Isolation of two highly active soybean (Glycine max (L.) Merr.) promoters and their characterization using a new automated image collection and analysis system

A novel automated image collection and analysis system was used to compare two new soybean (Glycine max (L.) Merr.) promoters with the cauliflower mosaic virus 35S (CaMV35S) promoter, which was used as an expression standard. For expression comparisons, various permutations of a soybean polyubiquitin (Gmubi) promoter, a soybean heat shock protein 90-like (GmHSP90L) promoter and the CaMV35S promoter were placed upstream of a green fluorescent protein (gfp) gene. DNA constructs were introduced via particle bombardment into excised cotyledons of germinating lima bean (Phaseolus lunatus L.) seeds, which were arranged in Petri dishes for automated image capture and image analysis. The automated system allowed monitoring and quantification of gfp gene expression in the same piece of tissue over time. The Gmubi promoter, with its intronic region intact, showed the highest expression that was over five times stronger than the CaMV35S promoter. When an intronic region was removed from the Gmubi promoter, GFP expression was reduced, but was still over two times greater than with the CaMV35S promoter. The full-length soybean GmHSP90L promoter was four times stronger than the CaMV35S promoter. Truncation of the GmHSP90L promoter resulted in stepwise decreases in promoter strength, which appear to correspond to removal of regulatory elements. Automated image capture and analysis allowed the rapid and efficient evaluation of these new promoters.

Injection of DNA into plant and animal tissues with micromechanical piercing structures

Silicon micromachining has been used to fabricate microprobes for injecting DNA into cells. Arrays of very sharp pyramidal points are etched on a silicon substrate. Pressing these points into a culture of cells, allows biologically active material to cross the cell wall barrier. Using the microprobes, DNA has been injected into plant (tobacco leaves) and animal (nematodes) cells.

MACHINE VISION EXTRACTED PLANT MOVEMENT FOR EARLY DETECTION OF PLANT WATER STRESS

A methodology was established for early, non-contact, and quantitative detection of plant water stress with machine vision extracted plant features. Top-projected canopy area (TPCA) of the plants was extracted from plant images using image-processing techniques. Water stress induced plant movement was decoupled from plant diurnal movement and plant growth using coefficient of relative variation of TPCA (CRV[TPCA)] and was found to be an effective marker for water stress detection. Threshold value of CRV(TPCA) as an indicator of water stress was determined by a parametric approach. The effectiveness of the sensing technique was evaluated against the timing of stress detection by an operator. Results of this study suggested that plant water stress detection using projected canopy area based features of the plants was feasible.

ESTABLISHING CROP WATER STRESS INDEX (CWSI) THRESHOLD VALUES FOR EARLY, NON-CONTACT DETECTION OF PLANT WATER STRESS

Early, non–contact, non–destructive, and quantitative detection of plant water stress with the application of infrared thermometry using a crop water stress index (CWSI) was established. A CWSI model for plants grown under controlled environments was developed using thermodynamic principles and energy balance of the plant. CWSI threshold values were established with a parametric approach. The effectiveness of the sensing technique was evaluated using timing of the stress detection by a grower. The CWSI–based technique was able to detect the stress one to two days prior to the time of stress detection by visual observation. Overall results of this study suggested that pre–visual and non–contact detection of plant water stress with infrared thermometry application using CWSI is feasible.

Sensing and End-Effector for a Robotic Tomato Harvester

Fresh produce is important for long-term space missions. It provides valuable nutritional needs and psychological boost for mission crews. Labor requirements to grow and harvest the crops, however, must be reduced through automation to allow the crew to perform other tasks. A robotic tomato harvester was developed for continuous, selective picking of mature tomatoes. The goal of this project was to develop a sensing unit and a robotic hand unit that could be integrated with a commercial robotic manipulator for the automated tomato harvesting task. Image processing algorithms were developed to determine sizes and locations of mature tomatoes including the ones that are partially occluded by leaves and/or branches. An end-effector subsystem, including a fourfinger prosthetic hand and an embedded hand controller, was designed and assembled for the tomato picking, holding and placing task. Improvement of a previously designed robotic hand resulted in a 50% weight reduction. The sensing and picking capability of the units has been demonstrated in laboratory and commercial greenhouse environments. Success rates of tomato fruit sensing and picking were better than 95% and 85%, respectively.

Monitoring of plant development in controlled environment with machine vision

Information acquisition is the foremost requirement for the control and continued operation of any complex system. This is especially true when a plant production system is used as a major component in a sustainable life support system. The plant production system not only provides food and fiber but is a means of providing critically needed life supporting elements such as O2 and purified H2O. The success of the plant production system relies on close monitoring and control of the production system. Machine vision technology was evaluated for the monitoring of plant health and development and showed promising results. Spectral and morphological characteristics of a model plant were studied under various artificially induced stress conditions. From the spectroscopic studies, it was found that the stresses can be determined from visual and non-visual symptoms. The development of the plant can also be quantified using a video image analysis base approach. The correlations between the qualities of the model plant and machine vision measured spectral features were established. The success of the research has shown a great potential in building an automated, closed-loop plant production system in controlled environments.

Root-zone temperature and nitrogen affect the yield and secondary metabolite concentration of fall- and spring-grown, high-density leaf lettuce.

BACKGROUND Understanding the effects of temperature and nitrogen levels on key variables, particularly under field conditions during cool seasons of temperate climates, is important. Here, we document the impact of root-zone heating and nitrogen (N) fertility on the accumulation and composition of fall- and spring-grown lettuce biomass. A novel, scalable field system was employed. RESULTS Direct-seeded plots containing a uniform, semi-solid, and nearly stable rooting medium were established outdoors in 2009 and 2010; each contained one of eight combinations of root-zone heating (-/+) and N fertility (0, 72, 144, and 576 mg day(-1)). Root-zone heating increased but withholding N decreased biomass accumulation in both years. Low N supplies were also associated with greater anthocyanin and total antioxidant power but lower N and phosphorus levels. Tissue chlorophyll a and vitamin C levels tracked root-zone temperature and N fertility more closely in 2009 and 2010, respectively. CONCLUSIONS Experimentally imposed root-zone temperature and N levels influenced the amount and properties of fall- and spring-grown lettuce tissue. Ambient conditions, however, dictated which of these factors exerted the greatest effect on the variables measured. Collectively, the results point to the potential for gains in system sustainability and productivity, including with respect to supplying human nutritional units.

Machine Vision Assisted Robotic Seedling Transplanting

A machine vision assisted robotic transplanter was developed to improve the quality of the grow-out trays that contain transplanted plugs. The machine vision system was designed to identify and locate empty positions in the transplanted seedling trays. The detection algorithm was able to inspect uneven height growth medium blocks in randomly oriented seedling trays. Fifteen species of seedlings were used in evaluating the integrated machine vision assisted robotic system. The overall vision system success rate in identifying mistransplanted empty growth medium blocks in the transplanted seedling trays is 95%.

Comparative analysis of 35S and lectin promoters in transgenic soybean tissue using an automated image acquisition system and image analysis

Expression of the green fluorescent protein (gfp) gene, under regulatory control of either the constitutive 35S promoter or the developmentally-regulated lectin promoter was monitored and quantified using a newly-developed automated tracking system. The automated system consisted of a computer-controlled two-dimensional robotics table and a programmable image acquisition system, which was used to semi-continuously monitor gfp gene expression during development of transgenic soybean [Glycine max (L.) Merrill] somatic embryos. Quantitative analysis of GFP expression showed that, during somatic embryo proliferation and early development, expression of lectin-GFP was not detected. During late embryo development, expression of lectin-GFP gradually increased until the levels were similar to those of 35S-GFP. The use of an automated image collection system and image analysis facilitated the frequent monitoring and quantification of gfp gene expression on a large number of samples over an extended period of time.