As the global population continues to grow, feeding the increasing number of people becomes a more pressing issue. The agricultural sector, responsible for feeding the world, is struggling to keep up with the demand. According to projections, the expected population of over 10 billion people in 2050 requires a two-fold increase in current food production. Unfortunately, the yields of the world’s staple food commodities are increasing at a rate far lower than the 2.4% per year required to achieve this goal. Compounding the problem is the steep labor costs and shrinking agricultural labor supply. To address these issues, increasing the yield efficiency is necessary. Machine vision has the potential to increase yield efficiency by enabling automated agricultural crop/fruit picking, ripeness detection, yield estimation, and fruit counting.
Yield Efficiency of Current Methods
Conventional agricultural crop/fruit picking, ripeness detection, yield estimation, and fruit counting techniques, such as mechanical harvesting and manual sampling, have been used for centuries, if not longer. These methods can be labor-intensive, time-consuming, and prone to human error, leading to decreased yield efficiencies.
Many commercially available depth sensors, which find use in machine vision systems in this industry, have a very limited framerate. For instance, the current version of the Kinect’s acquisition speed is limited to 30 frames per second. This makes it unsuitable for applications that require high throughput, like online quality inspection and phenotyping systems.
In contrast, Structured Light Based Machine Vision Systems offer a solution to these challenges. By enabling automated picking and counting, the system can increase yield efficiencies and reduce labor costs. The high speed and sub-millimeter accuracy of the system allow for precise yield estimation, which can help farmers make data-driven decisions to optimize their production processes. Ultimately, the structured light system has the potential to outperform traditional methods in terms of yield efficiency and accuracy, making it a promising solution for farmers looking to increase their productivity and profitability.
Improving Yield Efficiency using Fringe Projection Profilometry
One promising method to achieve increased yield efficiency is the use of a structured light-based machine vision system technique called Fringe Projection Profilometry (FPP). This technique, in comparison to other existing methods such as time of flight and laser, can perform depth imaging to sub-millimeter accuracy at high speeds (several thousand frames per second), allowing for accurate and fast inspection of produce on inspection lines, online crop phenotyping, and other applications in the agricultural industry. A group of researchers has demonstrated the potential utility of its superior fine-scale characterization, in agriculture, in their tutorial article. They demonstrate the accurate 3D measurement of a soybean root with a complex structure and, additionally, compare the fine-scale features of the depth map of a tree branch generated by both FPP and the time of flight-based Kinect v2.
Automatic crop harvesting machines possess a significant advantage over their human counterparts, owing to their ability to remain impervious to extreme temperatures and humidity and operate continuously around the clock without any hindrance. According to various works surveyed by a study published in Frontiers in Plant Science, existing stock machine vision systems have achieved picking success rates (defined as the overall percentage of crop successfully harvested) up to 88%, with the corresponding crop detection accuracies varying from 55% to 94%. Using FPP-based vision systems could lead to higher recognition accuracies and higher picking rates, which in turn can lead to higher overall yield efficiency.
Crop health monitoring is a continuous and tedious process, and large farms make crop inspection tedious and labor-intensive. This leads to a lot of produce spoiled by disease, which could have been preventable with timely chemical intervention. FPP-based machine vision systems could be utilized to achieve accurate round-the-clock crop health and disease monitoring, allowing farmers to implement precisely calibrated chemical treatments or alternative solutions, resulting in significantly less crop loss and improved yield efficiency.
In terms of yield estimation, traditional methods such as manual sampling and weighing can be time-consuming and may not provide accurate estimates of the entire crop. In contrast, using computer vision and machine learning algorithms on 3D image data collected by performing FPP can provide accurate yield estimations in real-time, allowing farmers to optimize their harvesting and production processes.
Overall, Structured Light Based Machine Vision Systems have the potential to significantly increase yield efficiencies compared to traditional methods, making them a promising solution for farmers and AgTech companies looking to improve their productivity and profitability.
Basic Setup of the FPP System
The foundation of a high-speed structured light-based vision system lies in its simple yet effective setup. It consists of three key components, a camera, a projector and an Arduino synchronization circuit. The projector projects black and white fringe patterns onto the object, while the camera acts as the eyes of the system and captures multiple images of the fringe-patterned object. The Arduino synchronization circuit ensures coordination between the camera and the projector, which is especially critical for high-speed pattern projection. The captured fringe-patterned images are then used to compute a 3D reconstruction of the object. Detailed steps and algorithms necessary for system calibration and 3D reconstruction are explained in the tutorial article authored by Badrinath Balasubramaniam, a student pursuing his doctorate in Mechanical Engineering in Dr. Beiwen Li’s laboratory at Iowa State University. Further, sample results outlined in the paper demonstrate the utility of this technique for agricultural applications.
Future Outlook: Improving the Structured Light System
To create an even more accurate depth measurement system that is faster and works based on the acquisition of a single image, a custom AI-based system trained on pertinent structured light image data can be used. This would improve the accuracy and speed of the system, making it an even more valuable tool for the agricultural industry.
Conclusion
Feeding the growing population of over 8 billion people is a daunting task, and it will only become more challenging as the population continues to increase. One solution to the problem of increasing yield efficiency is the use of structured light-based machine vision systems. These systems can perform depth imaging at high speeds and allow for fast and accurate inspection of produce, making them a valuable tool for the agricultural industry. With the potential to improve accuracy and speed through the use of custom AI-based systems, structured light-based machine vision systems are a promising solution to the challenge of feeding the world’s population.
About the Researcher
Badrinath is a dedicated Ph.D. student who specializes in the application of structured light and computer vision techniques to address complex engineering challenges. Currently, in the second year of pursuing his doctoral degree in Mechanical Engineering and Human-Computer Interaction at Iowa State University under the guidance of Dr. Beiwen Li in his laboratory, he is actively involved in cutting-edge research in this field. Originally from Chennai, India, Badrinath’s academic journey began with a Bachelor’s degree in Mechanical Engineering from the esteemed Birla Institute of Technology and Science, Pilani. Motivated to expand his knowledge and skills, he subsequently embarked on a Master’s degree in Mechanical Engineering at Iowa State University. After the successful completion of his Master’s program, Badrinath gained valuable industry experience as a project manager for three years. However, driven by his passion for research and desire to make significant academic contributions, he decided to transition back to academia and pursue his Ph.D.
