Michio Kise

Vision-based localization and mapping for an autonomous mower

This paper presents a vision-based localization and mapping algorithm for an autonomous mower. We divide the task for robotic mowing into two separate phases, a teaching phase and a mowing phase. During the teaching phase, the mower estimates the 3D positions of landmarks and defines a boundary in the lawn with an estimate of its own trajectory. During the mowing phase, the location of the mower is estimated using the landmark and boundary map acquired from the teaching phase. Of particular interest for our work is ensuring that the estimator for landmark mapping will not fail due to the nonlinearity of the system during the teaching phase. A nonlinear observer is designed with pseudo-measurements of each landmarks depth to prevent the map estimator from diverging. Simultaneously, the boundary is estimated with an EKF. Measurements taken from an omnidirectional camera, an IMU, and a ground speed sensor are used for the estimation. Numerical simulations and offline teaching phase experiments with our autonomous mower demonstrate the potential of our algorithm.

PVS: A system for large scale outdoor perception performance evaluation

This paper describes the motivation, design and implementation of a Perception Validation System (PVS), a system for measuring the outdoor perception performance of an autonomous vehicle. The PVS relies on using large amounts of real world data and ground truth information to quantify performance aspects such as the rate of false positive or false negative detections of an obstacle detection system. Our system relies on a relational database infrastructure to achieve a high degree of flexibility in the type of analyses it can support.

Omnidirectional-vision-based estimation for containment detection of a robotic mower

In this paper, we present an omnidirectional-vision-based localization and mapping system which can detect whether a robotic mower is contained in a permitted area. We exploit a robot-centric mapping framework that exploits a differential equation of motion of the landmarks, which are referenced with respect to the robot body frame. The estimator in our system generates a 3D point-based map with landmarks. Concurrently, the estimator defines a boundary of the mowing area with the estimated trajectory of the mower. The estimated boundary and the landmark map are provided for the estimation of the mowing location and for the containment detection. We validate the effectiveness of our system through numerical simulations and present the results of the outdoor experiment that we conducted with our robotic mower.

Portable Multispectral Imaging Instrument for Food Industry

The overall goal of this paper is to develop a cost effective portable multispectral imaging instrument that inspectors or plant personnel can use for quality control. Food safety and quality in food industry, especially small plant is ongoing problem. Identification and separation of contaminated food is very important to protect the consumer from a potential source of food poisoning. The current methods of inspecting food quality and safety, however, is mostly done by human visual observation with the criteria of color, consistency, and composition used for identification. Therefore, there is a need to develop instrumental methods which can reduce inspector fatigue, variability, insure continuous inspection, and provide a safe and high quality food supply for the consumer. The development of a portable multispectral instrument for contaminant detection is able to improve food safety and quality inspection in identifying, developing, and validating imaging technologies that are economically viable, especially for small food processing plants to help them meet food safety requirements. As an example of multispectral imaging application, we designed and fabricated a portable multispectral instrument to collect and analyze spectra for real-time contaminant detection for small poultry processing plant. Specifically, the procedure to develop a portable multispectral instrument including sensor design, fabrication, calibration, data collection, analysis, algorithm development for detecting poultry fecal contaminants in real-time was demonstrated.

Optical design and system calibration for three-band spectral imaging system with interchangeable filters

The design and calibration of a three-band image acquisition system was reported in this paper. The prototype system developed in this research was a three-band spectral imaging system that acquired two visible-band images and a NIR/IR image simultaneously. This was accomplished by using a three-port imaging system that consisted of three identical monochrome cameras, optical system, and three interchangeable optical filters. Spectral reflectance from an object was collimated by a front lens, and split in three ways by a cold mirror and beam-splitter: A cold mirror reflects 90% visible light and transmits 80% IR/NIR light. The visible light was again split identically in two directions by a beam-splitter. Focusing lenses then projected each image on its respective sensor. By incorporating an interchangeable filter design, the imaging system can measure any two visible spectral bands that range between 400 nm and 700 nm, and one NIR/IR band that ranges between 700 nm and 1000 nm without any complicated manufacturing process. In order to co-register the three-band images, a system-specific calibration algorithm was developed that compensates lens-sensor geometric misalignments.

Improving Performance of Real-time Multispectral Imaging System

A real-time multispectral imaging system has demonstrated a science-based tool for fecal and ingesta contaminant detection during poultry processing. In order to implement this imaging system at commercial poultry processing industry, the false positives must be removed. For doing this, we tested and implemented additional image processing algorithms including binning, cuticle removal filter, median filter, and morphological analysis in real-time mode to maximize detection accuracy and minimize false positives (FPs). The median filtering and binning process were able to reduce FPs up to 98.7% and 95.2%, respectively by eliminating most salt and pepper noise from the raw images. The detection accuracy varied with parameter values of image processing algorithms including binning, threshold, median filter, and morphological filter. Overall contaminant detection accuracy on moving birds varied from 84.3% to 97.8%. In this case, the FPs errors were 1.9% and 41.8%, respectively. Although neither the overall detection accuracy nor FPs errors were affected by camera gains, the results of detection accuracy were slightly changed from 87.4% to 95.1%. In this case, the FPs errors were 1.8% and 15.9%, respectively. Thus, the ARS multispectral imaging system was able to detect contaminants with 91.6% accuracy and 3.3% FPs errors by selecting optimum image processing methods at the processing speed of 140 birds per minute.

A Compact Multispectral Imaging System for Online Poultry Contaminant Inspection

The final goal of this research was to design and fabricate a compact, cost effective multispectral instrument for real-time contaminant detection at poultry processing plants. The prototype system developed in this research was a dual-band spectral imaging system that acquired two different spectral images simultaneously. It was a two-port imaging system that consisted of two identical monochrome cameras, optical system and two interchangeable optical filters. A spectral reflectance from an object was collimated by lenses and split identically in two directions by a beamsplitter, and then each light was focused on the sensor by lenses through an optical filter. Two optical filters, that determined the spectral characteristic of the imaging system, were interchangeable without complicated manufacturing process. To create an accurately registered two-band image, an image calibration algorithm that corrected lens distortions and lens-sensor geometric misalignments were developed.