Progress 09/01/16 to 08/31/21
Outputs
Target Audience:During the project development, we have reached california strawberry industries, growers, strawberry processors, and attending meetings and discussions in scientific communites. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?We have presented the technology to the variety of the audients, disseminated the work and reportedto the strawberry and food industry and scientific community and extensions. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
In summary, with our vigorous efforts, the project has achieved all the goals for an industrially viable automated system. All laboratory system tests, including algorithms, electronics, mechanical arms, waterjet arms, and motor controls, are completed . The engineering and manufacturing of the new prototype machine for the in-field testing are in progress. However, due to the covid19 pandemic shutdown, the manufacturing work had been halted. The process of constructing the new design of the system was stopped as a result. As a result, our system is still in the construction phase as of now. Except for the mechanical manufacturing work of the new prototype II machine, all other works have been accomplished: 1) Deep learning algorithms on calyx cutting and control system have been developed in its effectiveness and accuracy; 2) We have implemented smart waterjet cutting through a 128-point digital curved cutting which enables the cutting of strawberries in various orientations. 3) The system can handle more cases of strawberries that were rejected by previous designs and instances of unripe strawberries, berries with dirt, and green apex. 4) The lab test results showed around 92%-95% total accuracy of various situations. 5) Redesigned the mechanical motor control to improve the point-to-point movement time lag and improve the whole system processing speed ready for large throughput processing. We will continue to bring the project to complete fruition to ensure the system's reliability and durability in the field environments required by the strawberry processing industry. Our success will enhance the industry's ability to deliver quality, safe, and cost-effective fruit products to American consumers.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Wang, Dongyi, Robert Vinson, Maxwell Holmes, Gary Seibel, Xuemei Cheng, and Yang Tao. 2019. Vision Intelligence guided strawberry decalyx machineautomatic vision-guided intelligent decalyxing network (AVIDnet). In 2019 ASABE Annual International Meeting, American Society of Agricultural and Biological Engineers, 2019. Paper number: 1901905.
- Type:
Other
Status:
Published
Year Published:
2020
Citation:
Tao, Y. and D. Wang. 2020. Recent Advances of Artificial Intelligence Applications in Food. IUFoST Scientific Information Bulletins (SIB), 2020(9).
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Tao, Y. 2019. Application of AI in food processing Machine Intelligence Embedded in Food Processing Automation Systems. IFT: Leveraging Big Data and Artificial Intelligence for Ushering Innovations from Farm to Fork". IFT Annual Conference, New Orleans, June 2-4, 2019.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Tao, Y. 2019. Machine Vision Guided Food Manufacturing. IFT Advanced Food Manufacturing Program. USDA NIFA PI Annual Report Meeting (invited). June 2, 2019. New Orleans.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2018
Citation:
Holmes, M. R. Vinson, D. Wang, Y. Tao, G. Seibel. 2018. Food Process Automations for Increase Food System Sustainability. Global Health Poster. College of Agriculture and Natural Resources. Poster Presentation. College Park Student Union.
|
Progress 09/01/19 to 08/31/20
Outputs
Target Audience:
Nothing Reported
Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?
Nothing Reported
What do you plan to do during the next reporting period to accomplish the goals?1. Waiting for the new machine construction frame to be ready. 2. To assemble the new machine construction frame and test the new generation machine's performance on accuracy, speedand throughput with labortory samples. 2. Try to conduct a field trip test on the new machine in the coming spring and summer season.
Impacts
What was accomplished under these goals?
We are waiting for the new manufacturing framework of the machine to start engineering process of putting everything together and it requires time. We have ordered all the parts that needed for manufactoring the new machine. We had done all the laboratory system tests, including algorithm, electronics, mechanical arms, waterjet arms and motor controls during the past years and are in the progress of engineering and manufacturing the new prototype machine for the in-field testing in California. However, due to the covid19 pandemic shutdown, the manufacturing work that carried out in a industrial enviroment in Pensilvinia had been halted. The process of constructing the new design of the system was stopped as a result. Our system is still in the construction phase as of now. However, the peak strawberry harvesting window for 2020 field testing was missed as well because of the pandemic shutdown. As a result, we could finish what we planed to achieve during this year. Despite of the mechanic manufacturing work of the new prototype machine, all other works were ready: 1) deep learning imaging algorithm on calyx cutting and control system has developed in its effectiveness and accuracy; 2) We have implemented smart waterjet cutting through a 128-point digital curved cutting which enables the cutting of strawberries in various orientations. 3) The system can handle more cases of strawberries which were rejecting by previous design, and instances of unripe strawberries, berries with dirt and green apex. 4) The lab test results showed around 92%-95% total accuracy of various situations. 5) redesigned the mechanical motor control to improve the point-to-point movement time lag and improve the whole system processing speed ready for large throughput processing. We are waiting for the new manufacturing framework of the machine to start engineering process of putting everything together and it requires time. We filed a US utility patent for the machine in 2020. We plan to bring the machine in-field testing in the strawberry industry in California in 2021.
Publications
|
Progress 09/01/18 to 08/31/19
Outputs
Target Audience:
Nothing Reported
Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?
Nothing Reported
What do you plan to do during the next reporting period to accomplish the goals?1. To assemble the new machine construction frame and test the new generation machine's performance on accuracy, speed and throughput with labortory samples. 2. Try to conduct a field trip test on the new machine.
Impacts
What was accomplished under these goals?
Methods and algorithms were developed to enable and advance machine intelligence in the new prototype systems. Large sample tests were conducted in the industrial operating environment and received expected results and high throughput. Dramatic performance improvement was achieved in lab tests, by implementing data-driven deep neural networks and applying the models to online strawberry orientation detection (94.5% Rejection decision sensitivity), overall 91.8% of accurate cutting rate, and by successful design of the motor control system that led to high resolution movement control of waterjets. We have implemented smart waterjet cutting through a 128-point digital curved cutting which enables the cutting of strawberries in various orientations. The system now can handle more cases of strawberries which were rejecting by previous design, and also instances of unripe strawberries, berries with dirt and green apex. The lab test results showed around 92% total accuracy of various situations. The overall dimensions of the machine were changed to more flexible sizes in order to adapt to various customer requirements.
Publications
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2019
Citation:
Vision Intelligence guided strawberry decalyx machine-automatic vision-guided intelligent decalyxing network(AVIDnet), D. Wang, R.Vision, M.Holmes, G.Seibel, X.Cheng, Y.Tao., Paper presented at the 2019 ASABE Annual International Meeting, ASABE paper number:1901905
|
Progress 09/01/17 to 08/31/18
Outputs
Target Audience:
Nothing Reported
Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?The new 2nd generation prototype system is arranged to field test and training of how to use of the machine in Califorinia strawberry growers around early spring 2019. How have the results been disseminated to communities of interest?
Nothing Reported
What do you plan to do during the next reporting period to accomplish the goals?1. need a new mechanic model of the frame-structure design to make the overall dimensions of the machine be changed to more flexible sizes in order to adapt to various customer requirements. 2.Re-design and change the new motor control and communication protocol. 3. Further effort on improving the neural network design and the cutting algorithm performance
Impacts
What was accomplished under these goals?
In the 2nd year of the project,extensive research, design and experiments have been conducted toward the designing and training the neural networks and deploying the trained models to an online cutting algorithm. Based on the testing data collected during the first year of the Phase II project, a large volume of data analysis was conducted. More than 350,000 sample images were processed and analyzed offline, and 540,000 ground truth images were generated and utilized for training by the deep learning algorithms, to 1) classify the strawberries as cutable or non-cutable based on orientations and to 2) optimize curved cutting guideline positions for cutable strawberries. a new prototype structure was designed and tested on this new neural network based algorithm on samples. This 2nd generation prototype was designed to overcome the drawbacks of the 1st generation prototype as discovered in the field tests earlier. Dramatic performance improvement was achieved in lab tests, by implementing data-driven deep neural networks and applying the models to online strawberry orientation detection (94.1% of not cuttable accuracy rate), curved cutting (98% of accurate cutting rate), and by successful design of the electronic control system that led to high resolution movement control of waterjets.
Publications
|
Progress 09/01/16 to 08/31/17
Outputs
Target Audience:
Nothing Reported
Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?
Nothing Reported
What do you plan to do during the next reporting period to accomplish the goals?With our strong efforts, the project has gone very well toward its goal during the 1st year. In the 2nd year, we will make continued effort on algorithm development for improved machine intelligence for guiding the waterjet cutting, the online strawberry orientation detection, flexible tilted cutting, data-driven deep neural network implementation, and defect detection via combined NIR/VIS imaging. In hardware, we will add to the control mechanisms to improve the system capability of non-ideally oriented strawberries. We will conduct tests to examine the effectiveness of the neural network VGGNet, and investigate an efficient detection and sorting method, which will be suitable to be used in high speed processing environment.
Impacts
What was accomplished under these goals?
1) an initial prototype unit with user interface has been developed and tested. The prototype consists of subsystems of hardware control electronics, calyxes cutting line detection algorithm, high-pressure water cutting system, residual rejection system and controls. The machine was designed to adaptive to wash-down capabilities to meet the industrial operation requirements. In-plant tests were conducted in the previous fall season and the results have been further studied and analyzed for the system development. 2) a multi-dimensional dynamic cutting strategy has been explored and experimented to improve the cut robustness. Preliminary algorithms designed to guide the cutting and the post-cutting residue detection algorithms have been designed and analyzed. The algorithms were tested on large samples collected from the in-plant test for analysis. 3) large test samples and image processing results have been statistical analyzed. A deep-learning neural network module was designed to be trained by the samples. This neural network structure is designed to perform the defect detection and sorting. Additionally, the neural network is ready to be trained for strawberry orientation detection to improve the cutting decision, thereby improve the cutting effectiveness. A detailed technical progress report that contains proprietary and confidential information is submitted to the national program leader.
Publications
|
|