Development of a machine vision system for corn plant population, spacing and height measurement

Shrestha, Dev
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Journal Issue

A system was developed to measure the spatial variability of early stage plant population density, spacing and plant height. The Truncated Ellipsoidal (TE) method was developed to segment plants from background. A patch matching algorithm was developed to sequence for video frames of corn row videos. Algorithm performance was analyzed across three tillage treatments, three growth stages from V3 to V8, and three population densities varying from 27,000 to 81,500 plants/ha. Overall, the algorithm estimated the number of plants in 6.1 m crop row lengths with an RMSE of 2.1 plants. Following this encouraging result, a component-based software architecture was developed to automate site specific field data acquisition, processing, and geo-referenced plant parameter extraction. The architecture supported acquisition and processing of different data streams such as digital video or digital serial communications. Based on this architecture, early stage corn population estimation (ESCOPE) software was developed which grabbed pre-recorded digital video from a vehicle-mounted camera that was passed over corn rows and acquired GPS-NMEA strings which were modulated and recorded on the audio channel. Reusability and extensibility characteristics were demonstrated by adding a class to acquire images from the hard drive and also by deriving a new image analyzer class to extract an additional feature. For the crop height measurement, two different sensing approaches, stereo vision and ultrasonic, were investigated as candidate technologies for vehicle-based corn height sensors. For the stereo vision method, a chain code-based stereo correspondence technique was developed to determine the disparity in the stereo image pair. The ultrasonic sensor measured the distance to an object by detecting the time of flight of ultrasonic sound waves. A good correlation was found between the measured and estimated height using both stereo vision and the ultrasonic sensor. For the stereo vision sensor, r2 between the maximum plant height and estimated height was 0.76. For the ultrasonic sensor, r2 between the 25th percentile of the group height statistics and plant collar height was 0.75.

Agricultural and biosystems engineering, Agricultural engineering (Agricultural power and machinery), Agricultural power and machinery