Abstract
The quality of inflorescence is evaluated by a complicated algorithm which consists of many factors such as length, weight, relationship between leaves and flowers, color, flower position and so on. The quality evaluation is done by the human eye and the criteria depend on seller at auction. However, an objective evaluation for inflorescence is desirable. In this study, some experiments on the evaluation of a spray formation of spray-type chrysanthemum were done by neural network. The chrysanthemums were grown under different conditions regarding planting density, illumination, treatment of plant growth regulator to get various spray formations. In this paper, peduncle length, angle between the main stem and each peduncle and internode length were directly measured manually and were used as input parameters of the neural network. From the results evaluated by the neural network and an expert, the followings were found. 1. Various spray formations were obtained with different planting density and especially very different formation was obtained with the application of plant growth regulators. 2. The spray formation of chrysanthemum grown under the standard planting density was given excellent evaluation by the expert, while that treated with growth retardant had the lowest evaluation. 3. It was observed that the first evaluation by the expert was different from the second one. This implies that human evaluation can be affected by some unknown conditions. 4. Evaluation results by the neural network followed those by the expert adequately. The difference of the evaluated values between them was within ±1. 5. It was considered that the performance of the neural network for chrysanthemums could be improved by the addition of teaching data and input parameters.
